models_statistics.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The autoreload extension is already loaded. To reload it, use:\n",
      "  %reload_ext autoreload\n"
     ]
    }
   ],
   "source": [
    "%matplotlib notebook\n",
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "import load_files\n",
    "import models\n",
    "from load_files import *\n",
    "from models import *\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(7352, 128, 9) (7352, 1)\n",
      "(2947, 128, 9) (2947, 1)\n",
      "(7352, 128, 9) (7352, 6) (2947, 128, 9) (2947, 6)\n"
     ]
    }
   ],
   "source": [
    "x_train, y_train, x_test, y_test = load_dataset('../data/HAR/UCI_HAR_Dataset/')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0., 0., 0., 0., 1., 0.],\n",
       "       [0., 0., 0., 0., 1., 0.],\n",
       "       [0., 0., 0., 0., 1., 0.],\n",
       "       ...,\n",
       "       [0., 1., 0., 0., 0., 0.],\n",
       "       [0., 1., 0., 0., 0., 0.],\n",
       "       [0., 1., 0., 0., 0., 0.]], dtype=float32)"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(7352, 128, 9)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "n_timesteps, n_features, n_outputs = x_train.shape[1], x_train.shape[2], y_train.shape[1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "keras imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras import Model\n",
    "from keras.layers import Lambda, Input, Dropout, Flatten, LSTM, Concatenate, Bidirectional, Conv1D\n",
    "from keras import backend as K\n",
    "from keras.callbacks import TensorBoard\n",
    "from time import time\n",
    "from keras import optimizers\n",
    "from keras.callbacks import ReduceLROnPlateau, EarlyStopping\n",
    "from time import time"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model generators\n",
    "\n",
    "here we are going to analyse the performance of diffierent methods to classify the time series"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "here we define recall and auc metrics that are not implemented in keras\n",
    "\n",
    "references:\n",
    "\n",
    "https://stackoverflow.com/questions/41032551/how-to-compute-receiving-operating-characteristic-roc-and-auc-in-keras"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "import custom_metrics; from custom_metrics import as_keras_metric\n",
    "import tensorflow as tf\n",
    "\n",
    "auc_roc = as_keras_metric(tf.metrics.auc)\n",
    "recall = as_keras_metric(tf.metrics.recall)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Some callbacks\n",
    "\n",
    "we define some callbacks that we are going to implement during training\n",
    "\n",
    "https://stackoverflow.com/questions/50874596/how-to-detect-the-epoch-where-keras-earlystopping-occurred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "#lr_cb = ReduceLROnPlateau(monitor = 'val_loss', factor = 0.5, min_delta = 0.01, patience = 3, verbose = 1)\n",
    "#es_cb = EarlyStopping(monitor = 'val_loss', min_delta=0.01, patience = 10, verbose = 1, restore_best_weights = True)\n",
    "\n",
    "min_delta_val = 0.01\n",
    "lr_cb = ReduceLROnPlateau(monitor = 'val_auc', mode='max', \n",
    "                          factor = 0.5, min_delta = min_delta_val, patience = 3, verbose = 1)\n",
    "es_cb = EarlyStopping(monitor = 'val_auc', mode='max', \n",
    "                      min_delta=min_delta_val, patience = 10, verbose = 1, restore_best_weights = True)\n",
    "\n",
    "default_callbacks = [lr_cb, es_cb]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Training parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras import optimizers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "N_train = 3\n",
    "adam = optimizers.adam(lr=0.01)\n",
    "validation_split_on_training = 0.2\n",
    "epochs = 60\n",
    "batch_size = 250"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "import models; from models import *\n",
    "\n",
    "def generate_trained_models(model_type):\n",
    "    \n",
    "    print_summary_only_once = True\n",
    "    trained_models = []\n",
    "    trained_models_best_epoch = []\n",
    "    trained_models_stats = []\n",
    "    trained_models_time_taken = []\n",
    "\n",
    "    model_name_base = model_type\n",
    "\n",
    "    if model_type =='dense_2':\n",
    "        model_generator = dense_1d_2_model_generator\n",
    "    elif model_type =='LSTM_2':\n",
    "        model_generator = lstm_model_2_generator\n",
    "    elif model_type =='dense':\n",
    "        model_generator = dense_1d_model_generator\n",
    "    elif model_type =='LSTM':\n",
    "        model_generator = lstm_model_generator\n",
    "    elif model_type =='ens':\n",
    "        model_generator = hybrid_ens_generator\n",
    "    elif model_type =='dense_fc':\n",
    "        model_generator = dense_fully_connected_model_generator        \n",
    "    elif model_type =='conv_1d':\n",
    "        model_generator = conv_1d_model_generator                \n",
    "    else:\n",
    "        raise ValueError('No model type found')\n",
    "    \n",
    "        \n",
    "    for i in range(N_train):\n",
    "        tic = time()\n",
    "        model_name = model_name_base + '_' + str(i)\n",
    "        tensorboard = TensorBoard(log_dir=\"logs/{}\".format(model_name + '_' + str(time())))\n",
    "        callbacks_model = default_callbacks + [tensorboard]\n",
    "\n",
    "        # generate model\n",
    "        model_input, model_output , _ = model_generator(n_timesteps, n_features, n_outputs)\n",
    "        model = Model(model_input, model_output, name = model_name)\n",
    "\n",
    "        #compile model\n",
    "        model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy', auc_roc, recall])\n",
    "        if print_summary_only_once:\n",
    "            model.summary()\n",
    "            print_summary_only_once = False\n",
    "\n",
    "        # train model\n",
    "        model.fit(x_train, \n",
    "                  y_train, epochs=epochs, \n",
    "                  batch_size=batch_size, \n",
    "                  validation_split=validation_split_on_training,\n",
    "                  verbose=True,\n",
    "                  shuffle=True,\n",
    "                  callbacks = callbacks_model)     \n",
    "        trained_models.append(model)\n",
    "\n",
    "        # training time\n",
    "        training_time = time()-tic\n",
    "        trained_models_time_taken.append(training_time)\n",
    "        print('training time: {}s'.format(training_time))\n",
    "\n",
    "        # early stopping epoch\n",
    "        best_epoch = es_cb.stopped_epoch\n",
    "        trained_models_best_epoch.append(best_epoch)\n",
    "\n",
    "        #append best stat\n",
    "        best_stats = {}\n",
    "        for key in model.history.history.keys():\n",
    "            best_stats[key] = model.history.history[key][best_epoch]\n",
    "        trained_models_stats.append(best_stats)\n",
    "    \n",
    "    trained_models_stats = pd.DataFrame(trained_models_stats)\n",
    "    return trained_models, trained_models_stats, trained_models_time_taken, trained_models_best_epoch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_5 (InputLayer)         (None, 128, 9)            0         \n",
      "_________________________________________________________________\n",
      "lstm_7 (LSTM)                (None, 150)               96000     \n",
      "_________________________________________________________________\n",
      "dropout_5 (Dropout)          (None, 150)               0         \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 150)               22650     \n",
      "_________________________________________________________________\n",
      "lstm_out (Dense)             (None, 6)                 906       \n",
      "=================================================================\n",
      "Total params: 119,556\n",
      "Trainable params: 119,556\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Train on 5881 samples, validate on 1471 samples\n",
      "Epoch 1/60\n",
      "5881/5881 [==============================] - 8s 1ms/step - loss: 1.7962 - acc: 0.4166 - auc: 0.6337 - recall: 0.9532 - val_loss: 1.4970 - val_acc: 0.4691 - val_auc: 0.7610 - val_recall: 0.9882\n",
      "Epoch 2/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.4232 - acc: 0.5253 - auc: 0.7859 - recall: 0.9923 - val_loss: 1.2394 - val_acc: 0.6200 - val_auc: 0.8127 - val_recall: 0.9945\n",
      "Epoch 3/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 2.2208 - acc: 0.5482 - auc: 0.8249 - recall: 0.9885 - val_loss: 2.2631 - val_acc: 0.6621 - val_auc: 0.8245 - val_recall: 0.9751\n",
      "Epoch 4/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.2140 - acc: 0.5739 - auc: 0.8293 - recall: 0.9734 - val_loss: 1.1084 - val_acc: 0.6465 - val_auc: 0.8373 - val_recall: 0.9767\n",
      "Epoch 5/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.0182 - acc: 0.6186 - auc: 0.8470 - recall: 0.9793 - val_loss: 1.0625 - val_acc: 0.6744 - val_auc: 0.8561 - val_recall: 0.9811\n",
      "Epoch 6/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.9303 - acc: 0.6393 - auc: 0.8627 - recall: 0.9821 - val_loss: 0.8840 - val_acc: 0.6465 - val_auc: 0.8690 - val_recall: 0.9836\n",
      "Epoch 7/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.8152 - acc: 0.6742 - auc: 0.8744 - recall: 0.9848 - val_loss: 0.9139 - val_acc: 0.6852 - val_auc: 0.8795 - val_recall: 0.9859\n",
      "Epoch 8/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.7278 - acc: 0.6997 - auc: 0.8839 - recall: 0.9866 - val_loss: 0.7409 - val_acc: 0.7199 - val_auc: 0.8881 - val_recall: 0.9874\n",
      "Epoch 9/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.6810 - acc: 0.7018 - auc: 0.8920 - recall: 0.9881 - val_loss: 0.7723 - val_acc: 0.6676 - val_auc: 0.8955 - val_recall: 0.9888\n",
      "Epoch 10/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.6717 - acc: 0.7067 - auc: 0.8985 - recall: 0.9894 - val_loss: 0.7100 - val_acc: 0.7070 - val_auc: 0.9016 - val_recall: 0.9899\n",
      "Epoch 11/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.6274 - acc: 0.7259 - auc: 0.9045 - recall: 0.9904 - val_loss: 0.6893 - val_acc: 0.7403 - val_auc: 0.9074 - val_recall: 0.9908\n",
      "Epoch 12/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.5994 - acc: 0.7400 - auc: 0.9100 - recall: 0.9912 - val_loss: 0.6791 - val_acc: 0.7281 - val_auc: 0.9125 - val_recall: 0.9916\n",
      "Epoch 13/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.6266 - acc: 0.7444 - auc: 0.9148 - recall: 0.9918 - val_loss: 0.7921 - val_acc: 0.7661 - val_auc: 0.9170 - val_recall: 0.9919\n",
      "Epoch 14/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.1918 - acc: 0.7478 - auc: 0.9184 - recall: 0.9914 - val_loss: 1.5687 - val_acc: 0.7240 - val_auc: 0.9187 - val_recall: 0.9896\n",
      "Epoch 15/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.1540 - acc: 0.6575 - auc: 0.9174 - recall: 0.9888 - val_loss: 0.7830 - val_acc: 0.7437 - val_auc: 0.9179 - val_recall: 0.9888\n",
      "Epoch 16/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.6647 - acc: 0.7592 - auc: 0.9195 - recall: 0.9890 - val_loss: 0.8238 - val_acc: 0.6934 - val_auc: 0.9210 - val_recall: 0.9891\n",
      "Epoch 17/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.5913 - acc: 0.7558 - auc: 0.9220 - recall: 0.9894 - val_loss: 0.6575 - val_acc: 0.7757 - val_auc: 0.9235 - val_recall: 0.9897\n",
      "\n",
      "Epoch 00017: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 18/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.5333 - acc: 0.7876 - auc: 0.9250 - recall: 0.9900 - val_loss: 0.6492 - val_acc: 0.7886 - val_auc: 0.9265 - val_recall: 0.9902\n",
      "Epoch 19/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.5071 - acc: 0.8021 - auc: 0.9280 - recall: 0.9905 - val_loss: 0.6361 - val_acc: 0.7988 - val_auc: 0.9294 - val_recall: 0.9908\n",
      "Epoch 20/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4877 - acc: 0.8114 - auc: 0.9307 - recall: 0.9910 - val_loss: 0.6204 - val_acc: 0.8069 - val_auc: 0.9320 - val_recall: 0.9912\n",
      "Epoch 21/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4694 - acc: 0.8175 - auc: 0.9333 - recall: 0.9914 - val_loss: 0.6001 - val_acc: 0.8382 - val_auc: 0.9345 - val_recall: 0.9916\n",
      "Epoch 22/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.6255 - acc: 0.8162 - auc: 0.9357 - recall: 0.9917 - val_loss: 0.8997 - val_acc: 0.7804 - val_auc: 0.9364 - val_recall: 0.9915\n",
      "\n",
      "Epoch 00022: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 23/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.7090 - acc: 0.7968 - auc: 0.9370 - recall: 0.9914 - val_loss: 0.7633 - val_acc: 0.8124 - val_auc: 0.9377 - val_recall: 0.9913\n",
      "Epoch 24/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.5287 - acc: 0.7961 - auc: 0.9384 - recall: 0.9913 - val_loss: 0.5898 - val_acc: 0.8049 - val_auc: 0.9393 - val_recall: 0.9914\n",
      "Epoch 25/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4786 - acc: 0.8249 - auc: 0.9402 - recall: 0.9916 - val_loss: 0.6181 - val_acc: 0.8396 - val_auc: 0.9411 - val_recall: 0.9917\n",
      "Epoch 26/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.5109 - acc: 0.8179 - auc: 0.9419 - recall: 0.9918 - val_loss: 0.6756 - val_acc: 0.7288 - val_auc: 0.9426 - val_recall: 0.9919\n",
      "Epoch 27/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.5129 - acc: 0.7789 - auc: 0.9430 - recall: 0.9920 - val_loss: 0.6032 - val_acc: 0.7872 - val_auc: 0.9436 - val_recall: 0.9922\n",
      "Epoch 28/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4736 - acc: 0.8067 - auc: 0.9442 - recall: 0.9923 - val_loss: 0.5967 - val_acc: 0.7940 - val_auc: 0.9449 - val_recall: 0.9925\n",
      "\n",
      "Epoch 00028: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 29/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4654 - acc: 0.8126 - auc: 0.9455 - recall: 0.9926 - val_loss: 0.5923 - val_acc: 0.7974 - val_auc: 0.9461 - val_recall: 0.9927\n",
      "Epoch 30/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4506 - acc: 0.8226 - auc: 0.9467 - recall: 0.9928 - val_loss: 0.5882 - val_acc: 0.8029 - val_auc: 0.9473 - val_recall: 0.9930\n",
      "Epoch 31/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4506 - acc: 0.8279 - auc: 0.9479 - recall: 0.9931 - val_loss: 0.5847 - val_acc: 0.8056 - val_auc: 0.9484 - val_recall: 0.9932\n",
      "\n",
      "Epoch 00031: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\n",
      "Epoch 32/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4404 - acc: 0.8279 - auc: 0.9490 - recall: 0.9933 - val_loss: 0.5816 - val_acc: 0.8063 - val_auc: 0.9495 - val_recall: 0.9934\n",
      "Epoch 33/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4448 - acc: 0.8262 - auc: 0.9500 - recall: 0.9935 - val_loss: 0.5776 - val_acc: 0.8097 - val_auc: 0.9505 - val_recall: 0.9936\n",
      "Epoch 34/60\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4488 - acc: 0.8298 - auc: 0.9510 - recall: 0.9936 - val_loss: 0.5852 - val_acc: 0.8137 - val_auc: 0.9515 - val_recall: 0.9937\n",
      "Epoch 35/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4472 - acc: 0.8289 - auc: 0.9519 - recall: 0.9938 - val_loss: 0.5702 - val_acc: 0.8205 - val_auc: 0.9524 - val_recall: 0.9939\n",
      "Epoch 36/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4604 - acc: 0.8369 - auc: 0.9528 - recall: 0.9939 - val_loss: 0.5676 - val_acc: 0.8199 - val_auc: 0.9532 - val_recall: 0.9940\n",
      "Epoch 37/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4419 - acc: 0.8335 - auc: 0.9536 - recall: 0.9941 - val_loss: 0.5685 - val_acc: 0.8246 - val_auc: 0.9540 - val_recall: 0.9941\n",
      "\n",
      "Epoch 00037: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05.\n",
      "Epoch 38/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4410 - acc: 0.8363 - auc: 0.9543 - recall: 0.9942 - val_loss: 0.5670 - val_acc: 0.8287 - val_auc: 0.9547 - val_recall: 0.9942\n",
      "Epoch 39/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4405 - acc: 0.8320 - auc: 0.9550 - recall: 0.9943 - val_loss: 0.5650 - val_acc: 0.8300 - val_auc: 0.9554 - val_recall: 0.9944\n",
      "Epoch 40/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4366 - acc: 0.8369 - auc: 0.9557 - recall: 0.9944 - val_loss: 0.5625 - val_acc: 0.8273 - val_auc: 0.9561 - val_recall: 0.9945\n",
      "\n",
      "Epoch 00040: ReduceLROnPlateau reducing learning rate to 1.5625000742147677e-05.\n",
      "Epoch 41/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4443 - acc: 0.8402 - auc: 0.9564 - recall: 0.9945 - val_loss: 0.5727 - val_acc: 0.8294 - val_auc: 0.9567 - val_recall: 0.9946\n",
      "Epoch 42/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4330 - acc: 0.8366 - auc: 0.9570 - recall: 0.9946 - val_loss: 0.5717 - val_acc: 0.8280 - val_auc: 0.9573 - val_recall: 0.9947\n",
      "Epoch 43/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4398 - acc: 0.8357 - auc: 0.9576 - recall: 0.9947 - val_loss: 0.5710 - val_acc: 0.8307 - val_auc: 0.9579 - val_recall: 0.9948\n",
      "\n",
      "Epoch 00043: ReduceLROnPlateau reducing learning rate to 7.812500371073838e-06.\n",
      "Epoch 44/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4356 - acc: 0.8376 - auc: 0.9581 - recall: 0.9948 - val_loss: 0.5594 - val_acc: 0.8300 - val_auc: 0.9584 - val_recall: 0.9948\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00044: early stopping\n",
      "training time: 293.0284957885742s\n",
      "Train on 5881 samples, validate on 1471 samples\n",
      "Epoch 1/60\n",
      "5881/5881 [==============================] - 8s 1ms/step - loss: 1.5610 - acc: 0.4011 - auc: 0.6489 - recall: 0.9569 - val_loss: 1.3276 - val_acc: 0.5418 - val_auc: 0.7702 - val_recall: 0.9988\n",
      "Epoch 2/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 1.4608 - acc: 0.5535 - auc: 0.8036 - recall: 0.9985 - val_loss: 3.2944 - val_acc: 0.4779 - val_auc: 0.8152 - val_recall: 0.9864\n",
      "Epoch 3/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 1.9089 - acc: 0.5270 - auc: 0.8110 - recall: 0.9689 - val_loss: 1.1568 - val_acc: 0.5588 - val_auc: 0.8218 - val_recall: 0.9697\n",
      "Epoch 4/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 2.0437 - acc: 0.5732 - auc: 0.8300 - recall: 0.9681 - val_loss: 0.9462 - val_acc: 0.6193 - val_auc: 0.8341 - val_recall: 0.9623\n",
      "Epoch 5/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.3971 - acc: 0.6094 - auc: 0.8398 - recall: 0.9642 - val_loss: 0.7673 - val_acc: 0.6818 - val_auc: 0.8485 - val_recall: 0.9651\n",
      "Epoch 6/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.8452 - acc: 0.6382 - auc: 0.8564 - recall: 0.9681 - val_loss: 0.7693 - val_acc: 0.6526 - val_auc: 0.8643 - val_recall: 0.9709\n",
      "Epoch 7/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.6706 - acc: 0.6980 - auc: 0.8716 - recall: 0.9731 - val_loss: 0.7158 - val_acc: 0.6696 - val_auc: 0.8787 - val_recall: 0.9751\n",
      "Epoch 8/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.6294 - acc: 0.7026 - auc: 0.8843 - recall: 0.9768 - val_loss: 0.7936 - val_acc: 0.6703 - val_auc: 0.8894 - val_recall: 0.9782\n",
      "Epoch 9/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.7475 - acc: 0.7208 - auc: 0.8934 - recall: 0.9793 - val_loss: 1.5014 - val_acc: 0.6587 - val_auc: 0.8968 - val_recall: 0.9797\n",
      "Epoch 10/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 1.4847 - acc: 0.6744 - auc: 0.8986 - recall: 0.9787 - val_loss: 3.5390 - val_acc: 0.5255 - val_auc: 0.8973 - val_recall: 0.9751\n",
      "Epoch 11/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 2.8043 - acc: 0.6057 - auc: 0.8924 - recall: 0.9683 - val_loss: 0.8839 - val_acc: 0.7437 - val_auc: 0.8916 - val_recall: 0.9653\n",
      "\n",
      "Epoch 00011: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 12/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.6764 - acc: 0.6932 - auc: 0.8927 - recall: 0.9644 - val_loss: 0.7520 - val_acc: 0.7580 - val_auc: 0.8938 - val_recall: 0.9634\n",
      "Epoch 13/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 1.6147 - acc: 0.7242 - auc: 0.8950 - recall: 0.9626 - val_loss: 0.6567 - val_acc: 0.7893 - val_auc: 0.8965 - val_recall: 0.9620\n",
      "Epoch 14/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 1.3336 - acc: 0.7504 - auc: 0.8983 - recall: 0.9616 - val_loss: 0.6675 - val_acc: 0.7335 - val_auc: 0.9001 - val_recall: 0.9616\n",
      "Epoch 15/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.0385 - acc: 0.7759 - auc: 0.9022 - recall: 0.9620 - val_loss: 0.6100 - val_acc: 0.8144 - val_auc: 0.9042 - val_recall: 0.9622\n",
      "Epoch 16/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.9460 - acc: 0.8152 - auc: 0.9064 - recall: 0.9626 - val_loss: 0.5551 - val_acc: 0.8103 - val_auc: 0.9085 - val_recall: 0.9629\n",
      "Epoch 17/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.6877 - acc: 0.8385 - auc: 0.9108 - recall: 0.9633 - val_loss: 0.5200 - val_acc: 0.8049 - val_auc: 0.9132 - val_recall: 0.9640\n",
      "Epoch 18/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4320 - acc: 0.8522 - auc: 0.9157 - recall: 0.9649 - val_loss: 0.5805 - val_acc: 0.7675 - val_auc: 0.9179 - val_recall: 0.9658\n",
      "Epoch 19/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4541 - acc: 0.8111 - auc: 0.9198 - recall: 0.9666 - val_loss: 0.5543 - val_acc: 0.8504 - val_auc: 0.9215 - val_recall: 0.9675\n",
      "Epoch 20/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 1.7668 - acc: 0.7492 - auc: 0.9225 - recall: 0.9675 - val_loss: 2.0309 - val_acc: 0.7165 - val_auc: 0.9212 - val_recall: 0.9652\n",
      "\n",
      "Epoch 00020: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 21/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.6904 - acc: 0.8448 - auc: 0.9222 - recall: 0.9650 - val_loss: 0.4715 - val_acc: 0.8273 - val_auc: 0.9238 - val_recall: 0.9656\n",
      "Epoch 22/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.4443 - acc: 0.8687 - auc: 0.9255 - recall: 0.9663 - val_loss: 0.4706 - val_acc: 0.8307 - val_auc: 0.9271 - val_recall: 0.9669\n",
      "Epoch 23/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.5895 - acc: 0.8283 - auc: 0.9286 - recall: 0.9675 - val_loss: 1.0194 - val_acc: 0.7791 - val_auc: 0.9295 - val_recall: 0.9678\n",
      "Epoch 24/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.5413 - acc: 0.8473 - auc: 0.9303 - recall: 0.9680 - val_loss: 0.4797 - val_acc: 0.8307 - val_auc: 0.9316 - val_recall: 0.9686\n",
      "\n",
      "Epoch 00024: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 25/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.4309 - acc: 0.8726 - auc: 0.9329 - recall: 0.9691 - val_loss: 0.4725 - val_acc: 0.8409 - val_auc: 0.9341 - val_recall: 0.9696\n",
      "Epoch 26/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.4417 - acc: 0.8766 - auc: 0.9353 - recall: 0.9700 - val_loss: 0.4655 - val_acc: 0.8484 - val_auc: 0.9364 - val_recall: 0.9704\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 27/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.3937 - acc: 0.8815 - auc: 0.9375 - recall: 0.9709 - val_loss: 0.4696 - val_acc: 0.8464 - val_auc: 0.9386 - val_recall: 0.9713\n",
      "Epoch 28/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.3710 - acc: 0.8796 - auc: 0.9396 - recall: 0.9718 - val_loss: 0.4686 - val_acc: 0.8511 - val_auc: 0.9406 - val_recall: 0.9722\n",
      "\n",
      "Epoch 00028: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\n",
      "Epoch 29/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.3675 - acc: 0.8832 - auc: 0.9416 - recall: 0.9726 - val_loss: 0.4698 - val_acc: 0.8443 - val_auc: 0.9425 - val_recall: 0.9730\n",
      "Epoch 30/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.3985 - acc: 0.8789 - auc: 0.9433 - recall: 0.9733 - val_loss: 0.4692 - val_acc: 0.8504 - val_auc: 0.9442 - val_recall: 0.9737\n",
      "Epoch 31/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.3900 - acc: 0.8806 - auc: 0.9450 - recall: 0.9740 - val_loss: 0.4654 - val_acc: 0.8552 - val_auc: 0.9457 - val_recall: 0.9743\n",
      "Epoch 32/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.4200 - acc: 0.8725 - auc: 0.9465 - recall: 0.9746 - val_loss: 0.4994 - val_acc: 0.8076 - val_auc: 0.9471 - val_recall: 0.9749\n",
      "Epoch 33/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.4005 - acc: 0.8556 - auc: 0.9478 - recall: 0.9752 - val_loss: 0.4882 - val_acc: 0.8144 - val_auc: 0.9484 - val_recall: 0.9756\n",
      "\n",
      "Epoch 00033: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05.\n",
      "Epoch 34/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.3629 - acc: 0.8561 - auc: 0.9490 - recall: 0.9759 - val_loss: 0.4873 - val_acc: 0.8205 - val_auc: 0.9496 - val_recall: 0.9762\n",
      "Epoch 35/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.3512 - acc: 0.8670 - auc: 0.9502 - recall: 0.9765 - val_loss: 0.4842 - val_acc: 0.8280 - val_auc: 0.9508 - val_recall: 0.9768\n",
      "Epoch 36/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.4096 - acc: 0.8628 - auc: 0.9513 - recall: 0.9771 - val_loss: 0.4807 - val_acc: 0.8348 - val_auc: 0.9518 - val_recall: 0.9773\n",
      "\n",
      "Epoch 00036: ReduceLROnPlateau reducing learning rate to 1.5625000742147677e-05.\n",
      "Epoch 37/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.3622 - acc: 0.8706 - auc: 0.9523 - recall: 0.9776 - val_loss: 0.4794 - val_acc: 0.8375 - val_auc: 0.9528 - val_recall: 0.9778\n",
      "Epoch 38/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.3449 - acc: 0.8745 - auc: 0.9533 - recall: 0.9781 - val_loss: 0.4786 - val_acc: 0.8430 - val_auc: 0.9538 - val_recall: 0.9784\n",
      "Epoch 39/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.3831 - acc: 0.8698 - auc: 0.9543 - recall: 0.9786 - val_loss: 0.4779 - val_acc: 0.8450 - val_auc: 0.9547 - val_recall: 0.9788\n",
      "Epoch 40/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.3494 - acc: 0.8721 - auc: 0.9552 - recall: 0.9790 - val_loss: 0.4776 - val_acc: 0.8477 - val_auc: 0.9556 - val_recall: 0.9793\n",
      "Epoch 41/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.3606 - acc: 0.8740 - auc: 0.9560 - recall: 0.9795 - val_loss: 0.4771 - val_acc: 0.8484 - val_auc: 0.9564 - val_recall: 0.9797\n",
      "Epoch 42/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.3585 - acc: 0.8733 - auc: 0.9568 - recall: 0.9799 - val_loss: 0.4767 - val_acc: 0.8498 - val_auc: 0.9572 - val_recall: 0.9801\n",
      "\n",
      "Epoch 00042: ReduceLROnPlateau reducing learning rate to 7.812500371073838e-06.\n",
      "Epoch 43/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.3773 - acc: 0.8786 - auc: 0.9576 - recall: 0.9803 - val_loss: 0.4766 - val_acc: 0.8511 - val_auc: 0.9579 - val_recall: 0.9805\n",
      "Epoch 44/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.3455 - acc: 0.8755 - auc: 0.9583 - recall: 0.9807 - val_loss: 0.4765 - val_acc: 0.8511 - val_auc: 0.9587 - val_recall: 0.9808\n",
      "Epoch 45/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.3608 - acc: 0.8750 - auc: 0.9590 - recall: 0.9810 - val_loss: 0.4764 - val_acc: 0.8511 - val_auc: 0.9593 - val_recall: 0.9812\n",
      "\n",
      "Epoch 00045: ReduceLROnPlateau reducing learning rate to 3.906250185536919e-06.\n",
      "Epoch 46/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.3779 - acc: 0.8738 - auc: 0.9596 - recall: 0.9813 - val_loss: 0.4763 - val_acc: 0.8511 - val_auc: 0.9599 - val_recall: 0.9815\n",
      "Epoch 47/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.3524 - acc: 0.8764 - auc: 0.9603 - recall: 0.9817 - val_loss: 0.4762 - val_acc: 0.8511 - val_auc: 0.9606 - val_recall: 0.9818\n",
      "Epoch 48/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.3494 - acc: 0.8767 - auc: 0.9609 - recall: 0.9820 - val_loss: 0.4761 - val_acc: 0.8511 - val_auc: 0.9612 - val_recall: 0.9821\n",
      "\n",
      "Epoch 00048: ReduceLROnPlateau reducing learning rate to 1.9531250927684596e-06.\n",
      "Epoch 49/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.3764 - acc: 0.8733 - auc: 0.9614 - recall: 0.9823 - val_loss: 0.4761 - val_acc: 0.8511 - val_auc: 0.9617 - val_recall: 0.9824\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00049: early stopping\n",
      "training time: 329.1411781311035s\n",
      "Train on 5881 samples, validate on 1471 samples\n",
      "Epoch 1/60\n",
      "5881/5881 [==============================] - 8s 1ms/step - loss: 1.7984 - acc: 0.4299 - auc: 0.6958 - recall: 0.9507 - val_loss: 1.3126 - val_acc: 0.4772 - val_auc: 0.7922 - val_recall: 0.9844\n",
      "Epoch 2/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 1.4176 - acc: 0.5586 - auc: 0.8169 - recall: 0.9887 - val_loss: 2.1104 - val_acc: 0.4541 - val_auc: 0.8318 - val_recall: 0.9826\n",
      "Epoch 3/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.0927 - acc: 0.6115 - auc: 0.8421 - recall: 0.9801 - val_loss: 0.9118 - val_acc: 0.6710 - val_auc: 0.8572 - val_recall: 0.9828\n",
      "Epoch 4/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.8627 - acc: 0.6575 - auc: 0.8683 - recall: 0.9850 - val_loss: 0.8165 - val_acc: 0.6628 - val_auc: 0.8783 - val_recall: 0.9867\n",
      "Epoch 5/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.7195 - acc: 0.7023 - auc: 0.8864 - recall: 0.9882 - val_loss: 0.7455 - val_acc: 0.7097 - val_auc: 0.8939 - val_recall: 0.9894\n",
      "Epoch 6/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.6326 - acc: 0.7191 - auc: 0.8997 - recall: 0.9903 - val_loss: 0.7147 - val_acc: 0.7192 - val_auc: 0.9055 - val_recall: 0.9911\n",
      "Epoch 7/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.5621 - acc: 0.7584 - auc: 0.9103 - recall: 0.9918 - val_loss: 0.7455 - val_acc: 0.7593 - val_auc: 0.9151 - val_recall: 0.9924\n",
      "Epoch 8/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.5825 - acc: 0.7801 - auc: 0.9189 - recall: 0.9926 - val_loss: 0.6747 - val_acc: 0.7593 - val_auc: 0.9224 - val_recall: 0.9929\n",
      "Epoch 9/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.5013 - acc: 0.7944 - auc: 0.9256 - recall: 0.9933 - val_loss: 0.7014 - val_acc: 0.7315 - val_auc: 0.9288 - val_recall: 0.9937\n",
      "Epoch 10/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.5189 - acc: 0.8215 - auc: 0.9313 - recall: 0.9936 - val_loss: 0.6239 - val_acc: 0.8287 - val_auc: 0.9343 - val_recall: 0.9939\n",
      "Epoch 11/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.0119 - acc: 0.8204 - auc: 0.9366 - recall: 0.9936 - val_loss: 2.8343 - val_acc: 0.7043 - val_auc: 0.9363 - val_recall: 0.9905\n",
      "Epoch 12/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.0872 - acc: 0.7834 - auc: 0.9348 - recall: 0.9871 - val_loss: 0.5901 - val_acc: 0.7593 - val_auc: 0.9362 - val_recall: 0.9871\n",
      "Epoch 13/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4103 - acc: 0.8504 - auc: 0.9383 - recall: 0.9875 - val_loss: 0.6069 - val_acc: 0.8328 - val_auc: 0.9403 - val_recall: 0.9879\n",
      "\n",
      "Epoch 00013: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 14/60\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.3587 - acc: 0.7988 - auc: 0.9413 - recall: 0.9872 - val_loss: 1.9574 - val_acc: 0.7553 - val_auc: 0.9402 - val_recall: 0.9844\n",
      "Epoch 15/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.7421 - acc: 0.8522 - auc: 0.9409 - recall: 0.9835 - val_loss: 0.7738 - val_acc: 0.7831 - val_auc: 0.9420 - val_recall: 0.9835\n",
      "Epoch 16/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.3771 - acc: 0.8704 - auc: 0.9432 - recall: 0.9840 - val_loss: 0.5116 - val_acc: 0.8552 - val_auc: 0.9449 - val_recall: 0.9845\n",
      "Epoch 17/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.2796 - acc: 0.8961 - auc: 0.9465 - recall: 0.9850 - val_loss: 0.5321 - val_acc: 0.8742 - val_auc: 0.9481 - val_recall: 0.9854\n",
      "Epoch 18/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.2549 - acc: 0.9058 - auc: 0.9496 - recall: 0.9858 - val_loss: 0.5929 - val_acc: 0.8865 - val_auc: 0.9510 - val_recall: 0.9862\n",
      "Epoch 19/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.2462 - acc: 0.9180 - auc: 0.9523 - recall: 0.9865 - val_loss: 0.5429 - val_acc: 0.8980 - val_auc: 0.9536 - val_recall: 0.9868\n",
      "\n",
      "Epoch 00019: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 20/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.5142 - acc: 0.9138 - auc: 0.9546 - recall: 0.9868 - val_loss: 0.6966 - val_acc: 0.8994 - val_auc: 0.9555 - val_recall: 0.9866\n",
      "Epoch 21/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.6473 - acc: 0.9078 - auc: 0.9561 - recall: 0.9863 - val_loss: 0.7579 - val_acc: 0.8919 - val_auc: 0.9567 - val_recall: 0.9859\n",
      "Epoch 22/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4823 - acc: 0.9213 - auc: 0.9575 - recall: 0.9858 - val_loss: 0.6803 - val_acc: 0.8946 - val_auc: 0.9582 - val_recall: 0.9857\n",
      "Epoch 23/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.9298 - acc: 0.8944 - auc: 0.9587 - recall: 0.9854 - val_loss: 1.1070 - val_acc: 0.8770 - val_auc: 0.9586 - val_recall: 0.9845\n",
      "\n",
      "Epoch 00023: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 24/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.0631 - acc: 0.8912 - auc: 0.9587 - recall: 0.9838 - val_loss: 1.1129 - val_acc: 0.8804 - val_auc: 0.9586 - val_recall: 0.9830\n",
      "Epoch 25/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.0848 - acc: 0.8891 - auc: 0.9587 - recall: 0.9823 - val_loss: 0.9731 - val_acc: 0.8844 - val_auc: 0.9586 - val_recall: 0.9816\n",
      "Epoch 26/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.1612 - acc: 0.8876 - auc: 0.9587 - recall: 0.9810 - val_loss: 1.2995 - val_acc: 0.8668 - val_auc: 0.9585 - val_recall: 0.9801\n",
      "\n",
      "Epoch 00026: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\n",
      "Epoch 27/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.1815 - acc: 0.8862 - auc: 0.9583 - recall: 0.9793 - val_loss: 1.2142 - val_acc: 0.8763 - val_auc: 0.9582 - val_recall: 0.9786\n",
      "Epoch 28/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.2059 - acc: 0.8861 - auc: 0.9581 - recall: 0.9779 - val_loss: 1.1299 - val_acc: 0.8722 - val_auc: 0.9580 - val_recall: 0.9772\n",
      "Epoch 29/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.2642 - acc: 0.8803 - auc: 0.9580 - recall: 0.9766 - val_loss: 1.2017 - val_acc: 0.8688 - val_auc: 0.9578 - val_recall: 0.9758\n",
      "\n",
      "Epoch 00029: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05.\n",
      "Epoch 30/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 1.2180 - acc: 0.8847 - auc: 0.9577 - recall: 0.9751 - val_loss: 1.0335 - val_acc: 0.8770 - val_auc: 0.9576 - val_recall: 0.9745\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00030: early stopping\n",
      "training time: 204.52269887924194s\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_8 (InputLayer)            (None, 128, 9)       0                                            \n",
      "__________________________________________________________________________________________________\n",
      "lambda_1 (Lambda)               [(None, 128), (None, 0           input_8[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_8 (Dense)                 (None, 30)           3870        lambda_1[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_9 (Dense)                 (None, 30)           3870        lambda_1[0][1]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_10 (Dense)                (None, 30)           3870        lambda_1[0][2]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_11 (Dense)                (None, 30)           3870        lambda_1[0][3]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_12 (Dense)                (None, 30)           3870        lambda_1[0][4]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_13 (Dense)                (None, 30)           3870        lambda_1[0][5]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_14 (Dense)                (None, 30)           3870        lambda_1[0][6]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_15 (Dense)                (None, 30)           3870        lambda_1[0][7]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_16 (Dense)                (None, 30)           3870        lambda_1[0][8]                   \n",
      "__________________________________________________________________________________________________\n",
      "lambda_2 (Lambda)               (None, 30, 9)        0           dense_8[0][0]                    \n",
      "                                                                 dense_9[0][0]                    \n",
      "                                                                 dense_10[0][0]                   \n",
      "                                                                 dense_11[0][0]                   \n",
      "                                                                 dense_12[0][0]                   \n",
      "                                                                 dense_13[0][0]                   \n",
      "                                                                 dense_14[0][0]                   \n",
      "                                                                 dense_15[0][0]                   \n",
      "                                                                 dense_16[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dropout_8 (Dropout)             (None, 30, 9)        0           lambda_2[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "flatten_1 (Flatten)             (None, 270)          0           dropout_8[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_17 (Dense)                (None, 300)          81300       flatten_1[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dropout_9 (Dropout)             (None, 300)          0           dense_17[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_18 (Dense)                (None, 300)          90300       dropout_9[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_19 (Dense)                (None, 20)           6020        dense_18[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_2_out (Dense)             (None, 6)            126         dense_19[0][0]                   \n",
      "==================================================================================================\n",
      "Total params: 212,576\n",
      "Trainable params: 212,576\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 5881 samples, validate on 1471 samples\n",
      "Epoch 1/60\n",
      "5881/5881 [==============================] - 1s 230us/step - loss: 1.0672 - acc: 0.5707 - auc: 0.7505 - recall: 0.9575 - val_loss: 0.5765 - val_acc: 0.7825 - val_auc: 0.9027 - val_recall: 1.0000\n",
      "Epoch 2/60\n",
      "5881/5881 [==============================] - 0s 57us/step - loss: 0.4546 - acc: 0.8140 - auc: 0.9310 - recall: 1.0000 - val_loss: 0.4610 - val_acc: 0.8443 - val_auc: 0.9482 - val_recall: 1.0000\n",
      "Epoch 3/60\n",
      "5881/5881 [==============================] - 0s 53us/step - loss: 0.3040 - acc: 0.8876 - auc: 0.9581 - recall: 1.0000 - val_loss: 0.4041 - val_acc: 0.8668 - val_auc: 0.9654 - val_recall: 1.0000\n",
      "Epoch 4/60\n",
      "5881/5881 [==============================] - 0s 52us/step - loss: 0.2421 - acc: 0.9097 - auc: 0.9701 - recall: 1.0000 - val_loss: 0.3532 - val_acc: 0.8933 - val_auc: 0.9739 - val_recall: 1.0000\n",
      "Epoch 5/60\n",
      "5881/5881 [==============================] - 0s 51us/step - loss: 0.1989 - acc: 0.9218 - auc: 0.9767 - recall: 1.0000 - val_loss: 0.3333 - val_acc: 0.9123 - val_auc: 0.9790 - val_recall: 1.0000\n",
      "Epoch 6/60\n",
      "5881/5881 [==============================] - 0s 53us/step - loss: 0.1577 - acc: 0.9398 - auc: 0.9809 - recall: 1.0000 - val_loss: 0.3049 - val_acc: 0.9143 - val_auc: 0.9826 - val_recall: 1.0000\n",
      "Epoch 7/60\n",
      "5881/5881 [==============================] - 0s 53us/step - loss: 0.1403 - acc: 0.9446 - auc: 0.9839 - recall: 1.0000 - val_loss: 0.3189 - val_acc: 0.9116 - val_auc: 0.9850 - val_recall: 1.0000\n",
      "Epoch 8/60\n",
      "5881/5881 [==============================] - 0s 53us/step - loss: 0.1224 - acc: 0.9503 - auc: 0.9859 - recall: 1.0000 - val_loss: 0.2760 - val_acc: 0.9245 - val_auc: 0.9868 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00008: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 9/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.1092 - acc: 0.9553 - auc: 0.9876 - recall: 1.0000 - val_loss: 0.2887 - val_acc: 0.9245 - val_auc: 0.9882 - val_recall: 1.0000\n",
      "Epoch 10/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.1018 - acc: 0.9570 - auc: 0.9888 - recall: 1.0000 - val_loss: 0.2993 - val_acc: 0.9266 - val_auc: 0.9893 - val_recall: 1.0000\n",
      "Epoch 11/60\n",
      "5881/5881 [==============================] - 0s 55us/step - loss: 0.0953 - acc: 0.9623 - auc: 0.9897 - recall: 1.0000 - val_loss: 0.2960 - val_acc: 0.9279 - val_auc: 0.9901 - val_recall: 1.0000\n",
      "Epoch 12/60\n",
      "5881/5881 [==============================] - 0s 54us/step - loss: 0.0934 - acc: 0.9606 - auc: 0.9904 - recall: 1.0000 - val_loss: 0.2924 - val_acc: 0.9279 - val_auc: 0.9907 - val_recall: 1.0000\n",
      "Epoch 13/60\n",
      "5881/5881 [==============================] - 0s 53us/step - loss: 0.0891 - acc: 0.9611 - auc: 0.9910 - recall: 1.0000 - val_loss: 0.2995 - val_acc: 0.9279 - val_auc: 0.9912 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00013: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 14/60\n",
      "5881/5881 [==============================] - 0s 53us/step - loss: 0.0845 - acc: 0.9640 - auc: 0.9915 - recall: 1.0000 - val_loss: 0.2962 - val_acc: 0.9279 - val_auc: 0.9916 - val_recall: 1.0000\n",
      "Epoch 15/60\n",
      "5881/5881 [==============================] - 0s 54us/step - loss: 0.0833 - acc: 0.9634 - auc: 0.9918 - recall: 1.0000 - val_loss: 0.2975 - val_acc: 0.9300 - val_auc: 0.9920 - val_recall: 1.0000\n",
      "Epoch 16/60\n",
      "5881/5881 [==============================] - 0s 54us/step - loss: 0.0834 - acc: 0.9645 - auc: 0.9922 - recall: 1.0000 - val_loss: 0.3008 - val_acc: 0.9313 - val_auc: 0.9923 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00016: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 17/60\n",
      "5881/5881 [==============================] - 0s 53us/step - loss: 0.0797 - acc: 0.9653 - auc: 0.9925 - recall: 1.0000 - val_loss: 0.3085 - val_acc: 0.9279 - val_auc: 0.9926 - val_recall: 1.0000\n",
      "Epoch 18/60\n",
      "5881/5881 [==============================] - 0s 54us/step - loss: 0.0824 - acc: 0.9645 - auc: 0.9927 - recall: 1.0000 - val_loss: 0.3059 - val_acc: 0.9286 - val_auc: 0.9928 - val_recall: 1.0000\n",
      "Epoch 19/60\n",
      "5881/5881 [==============================] - 0s 54us/step - loss: 0.0798 - acc: 0.9651 - auc: 0.9929 - recall: 1.0000 - val_loss: 0.3072 - val_acc: 0.9300 - val_auc: 0.9930 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00019: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\n",
      "Epoch 20/60\n",
      "5881/5881 [==============================] - 0s 54us/step - loss: 0.0785 - acc: 0.9663 - auc: 0.9931 - recall: 1.0000 - val_loss: 0.3056 - val_acc: 0.9293 - val_auc: 0.9932 - val_recall: 1.0000\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00020: early stopping\n",
      "training time: 17.986271381378174s\n",
      "Train on 5881 samples, validate on 1471 samples\n",
      "Epoch 1/60\n",
      "5881/5881 [==============================] - 1s 238us/step - loss: 1.0522 - acc: 0.5630 - auc: 0.7559 - recall: 0.9575 - val_loss: 0.6562 - val_acc: 0.7913 - val_auc: 0.8957 - val_recall: 1.0000\n",
      "Epoch 2/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.4436 - acc: 0.8218 - auc: 0.9268 - recall: 1.0000 - val_loss: 0.4560 - val_acc: 0.8566 - val_auc: 0.9472 - val_recall: 1.0000\n",
      "Epoch 3/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.2858 - acc: 0.8934 - auc: 0.9574 - recall: 1.0000 - val_loss: 0.3935 - val_acc: 0.8770 - val_auc: 0.9655 - val_recall: 1.0000\n",
      "Epoch 4/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.2256 - acc: 0.9128 - auc: 0.9703 - recall: 1.0000 - val_loss: 0.3299 - val_acc: 0.9069 - val_auc: 0.9744 - val_recall: 1.0000\n",
      "Epoch 5/60\n",
      "5881/5881 [==============================] - 0s 55us/step - loss: 0.1789 - acc: 0.9335 - auc: 0.9773 - recall: 1.0000 - val_loss: 0.3061 - val_acc: 0.9171 - val_auc: 0.9798 - val_recall: 1.0000\n",
      "Epoch 6/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.1508 - acc: 0.9424 - auc: 0.9817 - recall: 1.0000 - val_loss: 0.3197 - val_acc: 0.9239 - val_auc: 0.9832 - val_recall: 1.0000\n",
      "Epoch 7/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.1414 - acc: 0.9442 - auc: 0.9844 - recall: 1.0000 - val_loss: 0.3144 - val_acc: 0.9259 - val_auc: 0.9854 - val_recall: 1.0000\n",
      "Epoch 8/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.1240 - acc: 0.9478 - auc: 0.9863 - recall: 1.0000 - val_loss: 0.3117 - val_acc: 0.9239 - val_auc: 0.9871 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00008: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 9/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.1066 - acc: 0.9578 - auc: 0.9878 - recall: 1.0000 - val_loss: 0.3022 - val_acc: 0.9232 - val_auc: 0.9884 - val_recall: 1.0000\n",
      "Epoch 10/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.0978 - acc: 0.9599 - auc: 0.9889 - recall: 1.0000 - val_loss: 0.3067 - val_acc: 0.9259 - val_auc: 0.9894 - val_recall: 1.0000\n",
      "Epoch 11/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.0909 - acc: 0.9623 - auc: 0.9898 - recall: 1.0000 - val_loss: 0.3180 - val_acc: 0.9245 - val_auc: 0.9902 - val_recall: 1.0000\n",
      "Epoch 12/60\n",
      "5881/5881 [==============================] - 0s 55us/step - loss: 0.0890 - acc: 0.9643 - auc: 0.9905 - recall: 1.0000 - val_loss: 0.3108 - val_acc: 0.9273 - val_auc: 0.9908 - val_recall: 1.0000\n",
      "Epoch 13/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.0884 - acc: 0.9640 - auc: 0.9911 - recall: 1.0000 - val_loss: 0.3271 - val_acc: 0.9225 - val_auc: 0.9913 - val_recall: 1.0000\n",
      "Epoch 14/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.0828 - acc: 0.9646 - auc: 0.9915 - recall: 1.0000 - val_loss: 0.3149 - val_acc: 0.9245 - val_auc: 0.9917 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00014: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 15/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.0825 - acc: 0.9640 - auc: 0.9919 - recall: 1.0000 - val_loss: 0.3144 - val_acc: 0.9252 - val_auc: 0.9921 - val_recall: 1.0000\n",
      "Epoch 16/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.0787 - acc: 0.9660 - auc: 0.9922 - recall: 1.0000 - val_loss: 0.3281 - val_acc: 0.9279 - val_auc: 0.9924 - val_recall: 1.0000\n",
      "Epoch 17/60\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.0807 - acc: 0.9641 - auc: 0.9925 - recall: 1.0000 - val_loss: 0.3174 - val_acc: 0.9252 - val_auc: 0.9927 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00017: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 18/60\n",
      "5881/5881 [==============================] - 0s 57us/step - loss: 0.0763 - acc: 0.9655 - auc: 0.9928 - recall: 1.0000 - val_loss: 0.3266 - val_acc: 0.9273 - val_auc: 0.9929 - val_recall: 1.0000\n",
      "Epoch 19/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.0774 - acc: 0.9650 - auc: 0.9930 - recall: 1.0000 - val_loss: 0.3251 - val_acc: 0.9259 - val_auc: 0.9931 - val_recall: 1.0000\n",
      "Epoch 20/60\n",
      "5881/5881 [==============================] - 0s 55us/step - loss: 0.0774 - acc: 0.9677 - auc: 0.9932 - recall: 1.0000 - val_loss: 0.3271 - val_acc: 0.9259 - val_auc: 0.9932 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00020: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\n",
      "Epoch 21/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.0763 - acc: 0.9650 - auc: 0.9933 - recall: 1.0000 - val_loss: 0.3311 - val_acc: 0.9273 - val_auc: 0.9934 - val_recall: 1.0000\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00021: early stopping\n",
      "training time: 19.37779450416565s\n",
      "Train on 5881 samples, validate on 1471 samples\n",
      "Epoch 1/60\n",
      "5881/5881 [==============================] - 1s 247us/step - loss: 0.9489 - acc: 0.6336 - auc: 0.7914 - recall: 0.9575 - val_loss: 0.5720 - val_acc: 0.8042 - val_auc: 0.9219 - val_recall: 1.0000\n",
      "Epoch 2/60\n",
      "5881/5881 [==============================] - 0s 55us/step - loss: 0.4254 - acc: 0.8262 - auc: 0.9431 - recall: 1.0000 - val_loss: 0.4623 - val_acc: 0.8552 - val_auc: 0.9562 - val_recall: 1.0000\n",
      "Epoch 3/60\n",
      "5881/5881 [==============================] - 0s 55us/step - loss: 0.2997 - acc: 0.8800 - auc: 0.9636 - recall: 1.0000 - val_loss: 0.3675 - val_acc: 0.8824 - val_auc: 0.9695 - val_recall: 1.0000\n",
      "Epoch 4/60\n",
      "5881/5881 [==============================] - 0s 55us/step - loss: 0.2335 - acc: 0.9097 - auc: 0.9735 - recall: 1.0000 - val_loss: 0.3464 - val_acc: 0.8960 - val_auc: 0.9767 - val_recall: 1.0000\n",
      "Epoch 5/60\n",
      "5881/5881 [==============================] - 0s 55us/step - loss: 0.1815 - acc: 0.9279 - auc: 0.9791 - recall: 1.0000 - val_loss: 0.3318 - val_acc: 0.9062 - val_auc: 0.9813 - val_recall: 1.0000\n",
      "Epoch 6/60\n",
      "5881/5881 [==============================] - 0s 55us/step - loss: 0.1488 - acc: 0.9403 - auc: 0.9829 - recall: 1.0000 - val_loss: 0.3051 - val_acc: 0.9137 - val_auc: 0.9843 - val_recall: 1.0000\n",
      "Epoch 7/60\n",
      "5881/5881 [==============================] - 0s 55us/step - loss: 0.1286 - acc: 0.9498 - auc: 0.9855 - recall: 1.0000 - val_loss: 0.3184 - val_acc: 0.9055 - val_auc: 0.9865 - val_recall: 1.0000\n",
      "Epoch 8/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.1156 - acc: 0.9537 - auc: 0.9873 - recall: 1.0000 - val_loss: 0.3222 - val_acc: 0.9150 - val_auc: 0.9880 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00008: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 9/60\n",
      "5881/5881 [==============================] - 0s 57us/step - loss: 0.1043 - acc: 0.9566 - auc: 0.9886 - recall: 1.0000 - val_loss: 0.3196 - val_acc: 0.9198 - val_auc: 0.9891 - val_recall: 1.0000\n",
      "Epoch 10/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.0955 - acc: 0.9612 - auc: 0.9896 - recall: 1.0000 - val_loss: 0.3158 - val_acc: 0.9218 - val_auc: 0.9900 - val_recall: 1.0000\n",
      "Epoch 11/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.0937 - acc: 0.9600 - auc: 0.9904 - recall: 1.0000 - val_loss: 0.3164 - val_acc: 0.9177 - val_auc: 0.9907 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00011: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 12/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.0905 - acc: 0.9621 - auc: 0.9910 - recall: 1.0000 - val_loss: 0.3121 - val_acc: 0.9191 - val_auc: 0.9912 - val_recall: 1.0000\n",
      "Epoch 13/60\n",
      "5881/5881 [==============================] - 0s 57us/step - loss: 0.0885 - acc: 0.9619 - auc: 0.9915 - recall: 1.0000 - val_loss: 0.3248 - val_acc: 0.9205 - val_auc: 0.9917 - val_recall: 1.0000\n",
      "Epoch 14/60\n",
      "5881/5881 [==============================] - 0s 57us/step - loss: 0.0859 - acc: 0.9628 - auc: 0.9918 - recall: 1.0000 - val_loss: 0.3185 - val_acc: 0.9232 - val_auc: 0.9920 - val_recall: 1.0000\n",
      "Epoch 15/60\n",
      "5881/5881 [==============================] - 0s 57us/step - loss: 0.0822 - acc: 0.9638 - auc: 0.9922 - recall: 1.0000 - val_loss: 0.3251 - val_acc: 0.9252 - val_auc: 0.9923 - val_recall: 1.0000\n",
      "Epoch 16/60\n",
      "5881/5881 [==============================] - 0s 58us/step - loss: 0.0827 - acc: 0.9640 - auc: 0.9925 - recall: 1.0000 - val_loss: 0.3232 - val_acc: 0.9259 - val_auc: 0.9926 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00016: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 17/60\n",
      "5881/5881 [==============================] - 0s 57us/step - loss: 0.0797 - acc: 0.9658 - auc: 0.9927 - recall: 1.0000 - val_loss: 0.3200 - val_acc: 0.9259 - val_auc: 0.9929 - val_recall: 1.0000\n",
      "Epoch 18/60\n",
      "5881/5881 [==============================] - 0s 57us/step - loss: 0.0799 - acc: 0.9631 - auc: 0.9930 - recall: 1.0000 - val_loss: 0.3231 - val_acc: 0.9245 - val_auc: 0.9931 - val_recall: 1.0000\n",
      "Epoch 19/60\n",
      "5881/5881 [==============================] - 0s 56us/step - loss: 0.0802 - acc: 0.9619 - auc: 0.9932 - recall: 1.0000 - val_loss: 0.3253 - val_acc: 0.9245 - val_auc: 0.9932 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00019: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\n",
      "Epoch 20/60\n",
      "5881/5881 [==============================] - 0s 55us/step - loss: 0.0795 - acc: 0.9631 - auc: 0.9933 - recall: 1.0000 - val_loss: 0.3256 - val_acc: 0.9245 - val_auc: 0.9934 - val_recall: 1.0000\n",
      "Epoch 21/60\n",
      "5881/5881 [==============================] - 0s 57us/step - loss: 0.0771 - acc: 0.9657 - auc: 0.9935 - recall: 1.0000 - val_loss: 0.3266 - val_acc: 0.9252 - val_auc: 0.9935 - val_recall: 1.0000\n",
      "Epoch 22/60\n",
      "5881/5881 [==============================] - 0s 57us/step - loss: 0.0767 - acc: 0.9660 - auc: 0.9936 - recall: 1.0000 - val_loss: 0.3273 - val_acc: 0.9245 - val_auc: 0.9936 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00022: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05.\n",
      "Epoch 23/60\n",
      "5881/5881 [==============================] - 0s 57us/step - loss: 0.0764 - acc: 0.9670 - auc: 0.9937 - recall: 1.0000 - val_loss: 0.3269 - val_acc: 0.9239 - val_auc: 0.9938 - val_recall: 1.0000\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00023: early stopping\n",
      "training time: 21.107561111450195s\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_11 (InputLayer)           (None, 128, 9)       0                                            \n",
      "__________________________________________________________________________________________________\n",
      "lambda_7 (Lambda)               [(None, 128), (None, 0           input_11[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_44 (Dense)                (None, 20)           2580        lambda_7[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_45 (Dense)                (None, 20)           2580        lambda_7[0][1]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_46 (Dense)                (None, 20)           2580        lambda_7[0][2]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_47 (Dense)                (None, 20)           2580        lambda_7[0][3]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_48 (Dense)                (None, 20)           2580        lambda_7[0][4]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_49 (Dense)                (None, 20)           2580        lambda_7[0][5]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_50 (Dense)                (None, 20)           2580        lambda_7[0][6]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_51 (Dense)                (None, 20)           2580        lambda_7[0][7]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_52 (Dense)                (None, 20)           2580        lambda_7[0][8]                   \n",
      "__________________________________________________________________________________________________\n",
      "lambda_8 (Lambda)               (None, 20, 9)        0           dense_44[0][0]                   \n",
      "                                                                 dense_45[0][0]                   \n",
      "                                                                 dense_46[0][0]                   \n",
      "                                                                 dense_47[0][0]                   \n",
      "                                                                 dense_48[0][0]                   \n",
      "                                                                 dense_49[0][0]                   \n",
      "                                                                 dense_50[0][0]                   \n",
      "                                                                 dense_51[0][0]                   \n",
      "                                                                 dense_52[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dropout_14 (Dropout)            (None, 20, 9)        0           lambda_8[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "flatten_4 (Flatten)             (None, 180)          0           dropout_14[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_53 (Dense)                (None, 250)          45250       flatten_4[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_54 (Dense)                (None, 20)           5020        dense_53[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_55 (Dense)                (None, 6)            126         dense_54[0][0]                   \n",
      "==================================================================================================\n",
      "Total params: 73,616\n",
      "Trainable params: 73,616\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 5881 samples, validate on 1471 samples\n",
      "Epoch 1/60\n",
      "5881/5881 [==============================] - 1s 241us/step - loss: 1.2196 - acc: 0.5278 - auc: 0.7142 - recall: 0.9575 - val_loss: 0.7363 - val_acc: 0.6594 - val_auc: 0.8762 - val_recall: 1.0000\n",
      "Epoch 2/60\n",
      "5881/5881 [==============================] - 0s 47us/step - loss: 0.5611 - acc: 0.7676 - auc: 0.9109 - recall: 1.0000 - val_loss: 0.5054 - val_acc: 0.8178 - val_auc: 0.9333 - val_recall: 1.0000\n",
      "Epoch 3/60\n",
      "5881/5881 [==============================] - 0s 49us/step - loss: 0.3658 - acc: 0.8594 - auc: 0.9458 - recall: 1.0000 - val_loss: 0.4145 - val_acc: 0.8776 - val_auc: 0.9552 - val_recall: 1.0000\n",
      "Epoch 4/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.2778 - acc: 0.8944 - auc: 0.9615 - recall: 1.0000 - val_loss: 0.3648 - val_acc: 0.8912 - val_auc: 0.9666 - val_recall: 1.0000\n",
      "Epoch 5/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.2408 - acc: 0.9121 - auc: 0.9702 - recall: 1.0000 - val_loss: 0.3411 - val_acc: 0.8939 - val_auc: 0.9733 - val_recall: 1.0000\n",
      "Epoch 6/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.2095 - acc: 0.9197 - auc: 0.9756 - recall: 1.0000 - val_loss: 0.3334 - val_acc: 0.8960 - val_auc: 0.9777 - val_recall: 1.0000\n",
      "Epoch 7/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1855 - acc: 0.9333 - auc: 0.9792 - recall: 1.0000 - val_loss: 0.3047 - val_acc: 0.9123 - val_auc: 0.9807 - val_recall: 1.0000\n",
      "Epoch 8/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1672 - acc: 0.9340 - auc: 0.9819 - recall: 1.0000 - val_loss: 0.2924 - val_acc: 0.9164 - val_auc: 0.9830 - val_recall: 1.0000\n",
      "Epoch 9/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1553 - acc: 0.9413 - auc: 0.9840 - recall: 1.0000 - val_loss: 0.3005 - val_acc: 0.9116 - val_auc: 0.9848 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00009: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 10/60\n",
      "5881/5881 [==============================] - 0s 51us/step - loss: 0.1384 - acc: 0.9454 - auc: 0.9855 - recall: 1.0000 - val_loss: 0.2916 - val_acc: 0.9150 - val_auc: 0.9862 - val_recall: 1.0000\n",
      "Epoch 11/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1375 - acc: 0.9473 - auc: 0.9868 - recall: 1.0000 - val_loss: 0.2865 - val_acc: 0.9171 - val_auc: 0.9873 - val_recall: 1.0000\n",
      "Epoch 12/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1307 - acc: 0.9483 - auc: 0.9878 - recall: 1.0000 - val_loss: 0.2806 - val_acc: 0.9218 - val_auc: 0.9883 - val_recall: 1.0000\n",
      "Epoch 13/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1275 - acc: 0.9492 - auc: 0.9887 - recall: 1.0000 - val_loss: 0.2838 - val_acc: 0.9150 - val_auc: 0.9890 - val_recall: 1.0000\n",
      "Epoch 14/60\n",
      "5881/5881 [==============================] - 0s 49us/step - loss: 0.1236 - acc: 0.9526 - auc: 0.9893 - recall: 1.0000 - val_loss: 0.2788 - val_acc: 0.9218 - val_auc: 0.9896 - val_recall: 1.0000\n",
      "Epoch 15/60\n",
      "5881/5881 [==============================] - 0s 49us/step - loss: 0.1188 - acc: 0.9536 - auc: 0.9899 - recall: 1.0000 - val_loss: 0.2756 - val_acc: 0.9232 - val_auc: 0.9902 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00015: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 16/60\n",
      "5881/5881 [==============================] - 0s 49us/step - loss: 0.1151 - acc: 0.9522 - auc: 0.9904 - recall: 1.0000 - val_loss: 0.2762 - val_acc: 0.9232 - val_auc: 0.9906 - val_recall: 1.0000\n",
      "Epoch 17/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1146 - acc: 0.9539 - auc: 0.9908 - recall: 1.0000 - val_loss: 0.2794 - val_acc: 0.9225 - val_auc: 0.9910 - val_recall: 1.0000\n",
      "Epoch 18/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1152 - acc: 0.9536 - auc: 0.9912 - recall: 1.0000 - val_loss: 0.2778 - val_acc: 0.9218 - val_auc: 0.9913 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00018: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 19/60\n",
      "5881/5881 [==============================] - 0s 49us/step - loss: 0.1135 - acc: 0.9549 - auc: 0.9915 - recall: 1.0000 - val_loss: 0.2765 - val_acc: 0.9232 - val_auc: 0.9916 - val_recall: 1.0000\n",
      "Epoch 20/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1072 - acc: 0.9560 - auc: 0.9918 - recall: 1.0000 - val_loss: 0.2777 - val_acc: 0.9232 - val_auc: 0.9919 - val_recall: 1.0000\n",
      "Epoch 21/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1089 - acc: 0.9548 - auc: 0.9920 - recall: 1.0000 - val_loss: 0.2759 - val_acc: 0.9259 - val_auc: 0.9921 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00021: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\n",
      "Epoch 22/60\n",
      "5881/5881 [==============================] - 0s 51us/step - loss: 0.1085 - acc: 0.9565 - auc: 0.9922 - recall: 1.0000 - val_loss: 0.2768 - val_acc: 0.9211 - val_auc: 0.9923 - val_recall: 1.0000\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00022: early stopping\n",
      "training time: 20.384784698486328s\n",
      "Train on 5881 samples, validate on 1471 samples\n",
      "Epoch 1/60\n",
      "5881/5881 [==============================] - 1s 247us/step - loss: 1.1528 - acc: 0.5778 - auc: 0.7480 - recall: 0.9575 - val_loss: 0.6658 - val_acc: 0.7519 - val_auc: 0.8962 - val_recall: 1.0000\n",
      "Epoch 2/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.4834 - acc: 0.8186 - auc: 0.9268 - recall: 1.0000 - val_loss: 0.5134 - val_acc: 0.8498 - val_auc: 0.9454 - val_recall: 1.0000\n",
      "Epoch 3/60\n",
      "5881/5881 [==============================] - 0s 49us/step - loss: 0.3163 - acc: 0.8779 - auc: 0.9555 - recall: 1.0000 - val_loss: 0.4380 - val_acc: 0.8538 - val_auc: 0.9632 - val_recall: 1.0000\n",
      "Epoch 4/60\n",
      "5881/5881 [==============================] - 0s 49us/step - loss: 0.2508 - acc: 0.9043 - auc: 0.9680 - recall: 1.0000 - val_loss: 0.3791 - val_acc: 0.8967 - val_auc: 0.9720 - val_recall: 1.0000\n",
      "Epoch 5/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.2146 - acc: 0.9254 - auc: 0.9749 - recall: 1.0000 - val_loss: 0.3627 - val_acc: 0.8892 - val_auc: 0.9773 - val_recall: 1.0000\n",
      "Epoch 6/60\n",
      "5881/5881 [==============================] - 0s 51us/step - loss: 0.1838 - acc: 0.9323 - auc: 0.9792 - recall: 1.0000 - val_loss: 0.3233 - val_acc: 0.9164 - val_auc: 0.9809 - val_recall: 1.0000\n",
      "Epoch 7/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1623 - acc: 0.9386 - auc: 0.9823 - recall: 1.0000 - val_loss: 0.3132 - val_acc: 0.9164 - val_auc: 0.9834 - val_recall: 1.0000\n",
      "Epoch 8/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1490 - acc: 0.9437 - auc: 0.9844 - recall: 1.0000 - val_loss: 0.3066 - val_acc: 0.9245 - val_auc: 0.9853 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00008: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 9/60\n",
      "5881/5881 [==============================] - 0s 51us/step - loss: 0.1376 - acc: 0.9471 - auc: 0.9861 - recall: 1.0000 - val_loss: 0.3046 - val_acc: 0.9205 - val_auc: 0.9867 - val_recall: 1.0000\n",
      "Epoch 10/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1308 - acc: 0.9498 - auc: 0.9873 - recall: 1.0000 - val_loss: 0.2991 - val_acc: 0.9205 - val_auc: 0.9879 - val_recall: 1.0000\n",
      "Epoch 11/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1287 - acc: 0.9524 - auc: 0.9883 - recall: 1.0000 - val_loss: 0.3023 - val_acc: 0.9198 - val_auc: 0.9887 - val_recall: 1.0000\n",
      "Epoch 12/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1220 - acc: 0.9546 - auc: 0.9891 - recall: 1.0000 - val_loss: 0.3047 - val_acc: 0.9164 - val_auc: 0.9895 - val_recall: 1.0000\n",
      "Epoch 13/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1195 - acc: 0.9543 - auc: 0.9898 - recall: 1.0000 - val_loss: 0.2980 - val_acc: 0.9211 - val_auc: 0.9901 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00013: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 14/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1160 - acc: 0.9534 - auc: 0.9903 - recall: 1.0000 - val_loss: 0.2931 - val_acc: 0.9239 - val_auc: 0.9906 - val_recall: 1.0000\n",
      "Epoch 15/60\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1111 - acc: 0.9573 - auc: 0.9908 - recall: 1.0000 - val_loss: 0.2933 - val_acc: 0.9239 - val_auc: 0.9910 - val_recall: 1.0000\n",
      "Epoch 16/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1109 - acc: 0.9568 - auc: 0.9912 - recall: 1.0000 - val_loss: 0.2889 - val_acc: 0.9259 - val_auc: 0.9914 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00016: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 17/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1077 - acc: 0.9570 - auc: 0.9915 - recall: 1.0000 - val_loss: 0.2911 - val_acc: 0.9252 - val_auc: 0.9917 - val_recall: 1.0000\n",
      "Epoch 18/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1127 - acc: 0.9539 - auc: 0.9918 - recall: 1.0000 - val_loss: 0.2873 - val_acc: 0.9266 - val_auc: 0.9919 - val_recall: 1.0000\n",
      "Epoch 19/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1063 - acc: 0.9583 - auc: 0.9920 - recall: 1.0000 - val_loss: 0.2885 - val_acc: 0.9266 - val_auc: 0.9922 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00019: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\n",
      "Epoch 20/60\n",
      "5881/5881 [==============================] - 0s 50us/step - loss: 0.1076 - acc: 0.9585 - auc: 0.9923 - recall: 1.0000 - val_loss: 0.2890 - val_acc: 0.9245 - val_auc: 0.9924 - val_recall: 1.0000\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00020: early stopping\n",
      "training time: 20.504687070846558s\n",
      "Train on 5881 samples, validate on 1471 samples\n",
      "Epoch 1/60\n",
      "5881/5881 [==============================] - 2s 256us/step - loss: 1.3022 - acc: 0.5338 - auc: 0.6745 - recall: 0.9575 - val_loss: 0.7196 - val_acc: 0.7791 - val_auc: 0.8728 - val_recall: 1.0000\n",
      "Epoch 2/60\n",
      "5881/5881 [==============================] - 0s 47us/step - loss: 0.5229 - acc: 0.8160 - auc: 0.9129 - recall: 1.0000 - val_loss: 0.5327 - val_acc: 0.8409 - val_auc: 0.9362 - val_recall: 1.0000\n",
      "Epoch 3/60\n",
      "5881/5881 [==============================] - 0s 46us/step - loss: 0.3209 - acc: 0.8743 - auc: 0.9486 - recall: 1.0000 - val_loss: 0.4679 - val_acc: 0.8606 - val_auc: 0.9579 - val_recall: 1.0000\n",
      "Epoch 4/60\n",
      "5881/5881 [==============================] - 0s 46us/step - loss: 0.2599 - acc: 0.8966 - auc: 0.9634 - recall: 1.0000 - val_loss: 0.4334 - val_acc: 0.8742 - val_auc: 0.9680 - val_recall: 1.0000\n",
      "Epoch 5/60\n",
      "5881/5881 [==============================] - 0s 46us/step - loss: 0.2140 - acc: 0.9167 - auc: 0.9713 - recall: 1.0000 - val_loss: 0.4393 - val_acc: 0.8688 - val_auc: 0.9741 - val_recall: 1.0000\n",
      "Epoch 6/60\n",
      "5881/5881 [==============================] - 0s 49us/step - loss: 0.1929 - acc: 0.9257 - auc: 0.9762 - recall: 1.0000 - val_loss: 0.4055 - val_acc: 0.8668 - val_auc: 0.9781 - val_recall: 1.0000\n",
      "Epoch 7/60\n",
      "5881/5881 [==============================] - 0s 47us/step - loss: 0.1747 - acc: 0.9328 - auc: 0.9797 - recall: 1.0000 - val_loss: 0.3898 - val_acc: 0.8838 - val_auc: 0.9810 - val_recall: 1.0000\n",
      "Epoch 8/60\n",
      "5881/5881 [==============================] - 0s 47us/step - loss: 0.1549 - acc: 0.9415 - auc: 0.9821 - recall: 1.0000 - val_loss: 0.3899 - val_acc: 0.8838 - val_auc: 0.9830 - val_recall: 1.0000\n",
      "Epoch 9/60\n",
      "5881/5881 [==============================] - 0s 47us/step - loss: 0.1457 - acc: 0.9405 - auc: 0.9839 - recall: 1.0000 - val_loss: 0.3754 - val_acc: 0.8906 - val_auc: 0.9846 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00009: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 10/60\n",
      "5881/5881 [==============================] - 0s 48us/step - loss: 0.1294 - acc: 0.9498 - auc: 0.9853 - recall: 1.0000 - val_loss: 0.3751 - val_acc: 0.8906 - val_auc: 0.9859 - val_recall: 1.0000\n",
      "Epoch 11/60\n",
      "5881/5881 [==============================] - 0s 48us/step - loss: 0.1223 - acc: 0.9532 - auc: 0.9865 - recall: 1.0000 - val_loss: 0.3849 - val_acc: 0.8926 - val_auc: 0.9870 - val_recall: 1.0000\n",
      "Epoch 12/60\n",
      "5881/5881 [==============================] - 0s 48us/step - loss: 0.1176 - acc: 0.9560 - auc: 0.9874 - recall: 1.0000 - val_loss: 0.3886 - val_acc: 0.8939 - val_auc: 0.9878 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00012: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 13/60\n",
      "5881/5881 [==============================] - 0s 48us/step - loss: 0.1143 - acc: 0.9541 - auc: 0.9881 - recall: 1.0000 - val_loss: 0.3827 - val_acc: 0.9001 - val_auc: 0.9885 - val_recall: 1.0000\n",
      "Epoch 14/60\n",
      "5881/5881 [==============================] - 0s 48us/step - loss: 0.1113 - acc: 0.9563 - auc: 0.9888 - recall: 1.0000 - val_loss: 0.3877 - val_acc: 0.8953 - val_auc: 0.9890 - val_recall: 1.0000\n",
      "Epoch 15/60\n",
      "5881/5881 [==============================] - 0s 48us/step - loss: 0.1110 - acc: 0.9563 - auc: 0.9893 - recall: 1.0000 - val_loss: 0.3946 - val_acc: 0.8967 - val_auc: 0.9895 - val_recall: 1.0000\n",
      "Epoch 16/60\n",
      "5881/5881 [==============================] - 0s 47us/step - loss: 0.1097 - acc: 0.9563 - auc: 0.9897 - recall: 1.0000 - val_loss: 0.3935 - val_acc: 0.8980 - val_auc: 0.9899 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00016: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 17/60\n",
      "5881/5881 [==============================] - 0s 48us/step - loss: 0.1096 - acc: 0.9544 - auc: 0.9901 - recall: 1.0000 - val_loss: 0.3925 - val_acc: 0.8980 - val_auc: 0.9902 - val_recall: 1.0000\n",
      "Epoch 18/60\n",
      "5881/5881 [==============================] - 0s 47us/step - loss: 0.1053 - acc: 0.9566 - auc: 0.9904 - recall: 1.0000 - val_loss: 0.3950 - val_acc: 0.8994 - val_auc: 0.9905 - val_recall: 1.0000\n",
      "Epoch 19/60\n",
      "5881/5881 [==============================] - 0s 47us/step - loss: 0.1059 - acc: 0.9578 - auc: 0.9907 - recall: 1.0000 - val_loss: 0.4017 - val_acc: 0.9001 - val_auc: 0.9908 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00019: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\n",
      "Epoch 20/60\n",
      "5881/5881 [==============================] - 0s 47us/step - loss: 0.1072 - acc: 0.9565 - auc: 0.9909 - recall: 1.0000 - val_loss: 0.3973 - val_acc: 0.9001 - val_auc: 0.9910 - val_recall: 1.0000\n",
      "Epoch 21/60\n",
      "5881/5881 [==============================] - 0s 49us/step - loss: 0.1044 - acc: 0.9558 - auc: 0.9911 - recall: 1.0000 - val_loss: 0.3973 - val_acc: 0.8994 - val_auc: 0.9912 - val_recall: 1.0000\n",
      "Epoch 22/60\n",
      "5881/5881 [==============================] - 0s 48us/step - loss: 0.1051 - acc: 0.9561 - auc: 0.9913 - recall: 1.0000 - val_loss: 0.4009 - val_acc: 0.9007 - val_auc: 0.9914 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00022: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05.\n",
      "Epoch 23/60\n",
      "5881/5881 [==============================] - 0s 47us/step - loss: 0.1019 - acc: 0.9566 - auc: 0.9915 - recall: 1.0000 - val_loss: 0.4000 - val_acc: 0.9001 - val_auc: 0.9915 - val_recall: 1.0000\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00023: early stopping\n",
      "training time: 22.53015947341919s\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_14 (InputLayer)           (None, 128, 9)       0                                            \n",
      "__________________________________________________________________________________________________\n",
      "lambda_13 (Lambda)              [(None, 128), (None, 0           input_14[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_80 (Dense)                (None, 20)           2580        lambda_13[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_81 (Dense)                (None, 20)           2580        lambda_13[0][1]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_82 (Dense)                (None, 20)           2580        lambda_13[0][2]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_83 (Dense)                (None, 20)           2580        lambda_13[0][3]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_84 (Dense)                (None, 20)           2580        lambda_13[0][4]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_85 (Dense)                (None, 20)           2580        lambda_13[0][5]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_86 (Dense)                (None, 20)           2580        lambda_13[0][6]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_87 (Dense)                (None, 20)           2580        lambda_13[0][7]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_88 (Dense)                (None, 20)           2580        lambda_13[0][8]                  \n",
      "__________________________________________________________________________________________________\n",
      "lambda_14 (Lambda)              (None, 20, 9)        0           dense_80[0][0]                   \n",
      "                                                                 dense_81[0][0]                   \n",
      "                                                                 dense_82[0][0]                   \n",
      "                                                                 dense_83[0][0]                   \n",
      "                                                                 dense_84[0][0]                   \n",
      "                                                                 dense_85[0][0]                   \n",
      "                                                                 dense_86[0][0]                   \n",
      "                                                                 dense_87[0][0]                   \n",
      "                                                                 dense_88[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "lstm_10 (LSTM)                  (None, 150)          96000       input_14[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dropout_17 (Dropout)            (None, 20, 9)        0           lambda_14[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dropout_18 (Dropout)            (None, 150)          0           lstm_10[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "flatten_7 (Flatten)             (None, 180)          0           dropout_17[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_92 (Dense)                (None, 150)          22650       dropout_18[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_89 (Dense)                (None, 250)          45250       flatten_7[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_1 (Concatenate)     (None, 400)          0           dense_92[0][0]                   \n",
      "                                                                 dense_89[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_93 (Dense)                (None, 6)            2406        concatenate_1[0][0]              \n",
      "==================================================================================================\n",
      "Total params: 189,526\n",
      "Trainable params: 189,526\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 5881 samples, validate on 1471 samples\n",
      "Epoch 1/60\n",
      "5881/5881 [==============================] - 8s 1ms/step - loss: 1.3059 - acc: 0.5681 - auc: 0.7149 - recall: 0.9563 - val_loss: 0.7424 - val_acc: 0.7492 - val_auc: 0.8863 - val_recall: 0.9941\n",
      "Epoch 2/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 1.4624 - acc: 0.7665 - auc: 0.9065 - recall: 0.9835 - val_loss: 2.2352 - val_acc: 0.7308 - val_auc: 0.9088 - val_recall: 0.9660\n",
      "Epoch 3/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.7670 - acc: 0.8473 - auc: 0.9171 - recall: 0.9633 - val_loss: 2.0506 - val_acc: 0.7349 - val_auc: 0.9267 - val_recall: 0.9643\n",
      "Epoch 4/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.4437 - acc: 0.8898 - auc: 0.9318 - recall: 0.9637 - val_loss: 0.4475 - val_acc: 0.8559 - val_auc: 0.9405 - val_recall: 0.9680\n",
      "Epoch 5/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.2651 - acc: 0.9146 - auc: 0.9468 - recall: 0.9713 - val_loss: 0.4139 - val_acc: 0.8668 - val_auc: 0.9522 - val_recall: 0.9741\n",
      "Epoch 6/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.2299 - acc: 0.9279 - auc: 0.9564 - recall: 0.9762 - val_loss: 0.4037 - val_acc: 0.8736 - val_auc: 0.9601 - val_recall: 0.9782\n",
      "Epoch 7/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.2126 - acc: 0.9352 - auc: 0.9631 - recall: 0.9797 - val_loss: 0.3933 - val_acc: 0.8804 - val_auc: 0.9656 - val_recall: 0.9810\n",
      "Epoch 8/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1956 - acc: 0.9395 - auc: 0.9678 - recall: 0.9822 - val_loss: 0.3770 - val_acc: 0.8892 - val_auc: 0.9697 - val_recall: 0.9832\n",
      "Epoch 9/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1640 - acc: 0.9390 - auc: 0.9714 - recall: 0.9842 - val_loss: 0.3628 - val_acc: 0.8885 - val_auc: 0.9729 - val_recall: 0.9850\n",
      "Epoch 10/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.1511 - acc: 0.9452 - auc: 0.9742 - recall: 0.9858 - val_loss: 0.3504 - val_acc: 0.8939 - val_auc: 0.9755 - val_recall: 0.9865\n",
      "\n",
      "Epoch 00010: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 11/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1336 - acc: 0.9519 - auc: 0.9766 - recall: 0.9872 - val_loss: 0.3304 - val_acc: 0.9014 - val_auc: 0.9776 - val_recall: 0.9878\n",
      "Epoch 12/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1281 - acc: 0.9524 - auc: 0.9785 - recall: 0.9883 - val_loss: 0.3189 - val_acc: 0.9001 - val_auc: 0.9794 - val_recall: 0.9888\n",
      "Epoch 13/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.1505 - acc: 0.9534 - auc: 0.9801 - recall: 0.9892 - val_loss: 0.3786 - val_acc: 0.8980 - val_auc: 0.9807 - val_recall: 0.9895\n",
      "Epoch 14/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1902 - acc: 0.9517 - auc: 0.9812 - recall: 0.9896 - val_loss: 0.3214 - val_acc: 0.9035 - val_auc: 0.9817 - val_recall: 0.9900\n",
      "\n",
      "Epoch 00014: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 15/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1156 - acc: 0.9575 - auc: 0.9823 - recall: 0.9903 - val_loss: 0.3251 - val_acc: 0.9014 - val_auc: 0.9828 - val_recall: 0.9906\n",
      "Epoch 16/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1126 - acc: 0.9561 - auc: 0.9833 - recall: 0.9909 - val_loss: 0.3209 - val_acc: 0.9028 - val_auc: 0.9837 - val_recall: 0.9912\n",
      "Epoch 17/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1116 - acc: 0.9541 - auc: 0.9842 - recall: 0.9915 - val_loss: 0.3164 - val_acc: 0.9021 - val_auc: 0.9846 - val_recall: 0.9918\n",
      "\n",
      "Epoch 00017: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 18/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1094 - acc: 0.9570 - auc: 0.9849 - recall: 0.9920 - val_loss: 0.3193 - val_acc: 0.9048 - val_auc: 0.9853 - val_recall: 0.9922\n",
      "Epoch 19/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1104 - acc: 0.9583 - auc: 0.9856 - recall: 0.9924 - val_loss: 0.3211 - val_acc: 0.9007 - val_auc: 0.9859 - val_recall: 0.9926\n",
      "Epoch 20/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1098 - acc: 0.9554 - auc: 0.9862 - recall: 0.9928 - val_loss: 0.3184 - val_acc: 0.9035 - val_auc: 0.9865 - val_recall: 0.9930\n",
      "\n",
      "Epoch 00020: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\n",
      "Epoch 21/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.1067 - acc: 0.9582 - auc: 0.9867 - recall: 0.9932 - val_loss: 0.3194 - val_acc: 0.9028 - val_auc: 0.9870 - val_recall: 0.9933\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00021: early stopping\n",
      "training time: 155.91726851463318s\n",
      "Train on 5881 samples, validate on 1471 samples\n",
      "Epoch 1/60\n",
      "5881/5881 [==============================] - 9s 1ms/step - loss: 1.4477 - acc: 0.6195 - auc: 0.8052 - recall: 0.9534 - val_loss: 0.6318 - val_acc: 0.8171 - val_auc: 0.8996 - val_recall: 0.9779\n",
      "Epoch 2/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.4665 - acc: 0.8250 - auc: 0.9280 - recall: 0.9855 - val_loss: 0.5011 - val_acc: 0.8491 - val_auc: 0.9460 - val_recall: 0.9897\n",
      "Epoch 3/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.3281 - acc: 0.8750 - auc: 0.9550 - recall: 0.9918 - val_loss: 0.3967 - val_acc: 0.8756 - val_auc: 0.9620 - val_recall: 0.9933\n",
      "Epoch 4/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.2449 - acc: 0.9157 - auc: 0.9669 - recall: 0.9942 - val_loss: 0.3406 - val_acc: 0.8912 - val_auc: 0.9712 - val_recall: 0.9950\n",
      "Epoch 5/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.2180 - acc: 0.9264 - auc: 0.9742 - recall: 0.9955 - val_loss: 0.4218 - val_acc: 0.8967 - val_auc: 0.9766 - val_recall: 0.9958\n",
      "Epoch 6/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.2228 - acc: 0.9322 - auc: 0.9783 - recall: 0.9958 - val_loss: 0.3494 - val_acc: 0.9021 - val_auc: 0.9799 - val_recall: 0.9960\n",
      "Epoch 7/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.2012 - acc: 0.9383 - auc: 0.9812 - recall: 0.9961 - val_loss: 0.3625 - val_acc: 0.8987 - val_auc: 0.9822 - val_recall: 0.9962\n",
      "Epoch 8/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.2120 - acc: 0.9424 - auc: 0.9831 - recall: 0.9962 - val_loss: 0.4353 - val_acc: 0.9075 - val_auc: 0.9838 - val_recall: 0.9962\n",
      "\n",
      "Epoch 00008: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 9/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.1902 - acc: 0.9424 - auc: 0.9844 - recall: 0.9960 - val_loss: 0.2903 - val_acc: 0.9089 - val_auc: 0.9851 - val_recall: 0.9962\n",
      "Epoch 10/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1341 - acc: 0.9492 - auc: 0.9858 - recall: 0.9964 - val_loss: 0.2922 - val_acc: 0.9171 - val_auc: 0.9864 - val_recall: 0.9965\n",
      "Epoch 11/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.1259 - acc: 0.9510 - auc: 0.9869 - recall: 0.9967 - val_loss: 0.3025 - val_acc: 0.9116 - val_auc: 0.9874 - val_recall: 0.9969\n",
      "Epoch 12/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.1245 - acc: 0.9541 - auc: 0.9878 - recall: 0.9970 - val_loss: 0.2939 - val_acc: 0.9143 - val_auc: 0.9883 - val_recall: 0.9971\n",
      "Epoch 13/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.1196 - acc: 0.9551 - auc: 0.9886 - recall: 0.9972 - val_loss: 0.2918 - val_acc: 0.9205 - val_auc: 0.9890 - val_recall: 0.9973\n",
      "Epoch 14/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1144 - acc: 0.9556 - auc: 0.9893 - recall: 0.9974 - val_loss: 0.2932 - val_acc: 0.9177 - val_auc: 0.9896 - val_recall: 0.9975\n",
      "\n",
      "Epoch 00014: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 15/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.1149 - acc: 0.9548 - auc: 0.9898 - recall: 0.9976 - val_loss: 0.3012 - val_acc: 0.9116 - val_auc: 0.9901 - val_recall: 0.9977\n",
      "Epoch 16/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.1113 - acc: 0.9568 - auc: 0.9903 - recall: 0.9978 - val_loss: 0.2938 - val_acc: 0.9143 - val_auc: 0.9905 - val_recall: 0.9979\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 17/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.1097 - acc: 0.9573 - auc: 0.9907 - recall: 0.9979 - val_loss: 0.2921 - val_acc: 0.9089 - val_auc: 0.9909 - val_recall: 0.9980\n",
      "\n",
      "Epoch 00017: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 18/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1106 - acc: 0.9589 - auc: 0.9911 - recall: 0.9980 - val_loss: 0.3015 - val_acc: 0.9089 - val_auc: 0.9912 - val_recall: 0.9981\n",
      "Epoch 19/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.1079 - acc: 0.9568 - auc: 0.9914 - recall: 0.9981 - val_loss: 0.2955 - val_acc: 0.9157 - val_auc: 0.9915 - val_recall: 0.9982\n",
      "Epoch 20/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1070 - acc: 0.9568 - auc: 0.9916 - recall: 0.9982 - val_loss: 0.2934 - val_acc: 0.9164 - val_auc: 0.9918 - val_recall: 0.9983\n",
      "\n",
      "Epoch 00020: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\n",
      "Epoch 21/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1076 - acc: 0.9573 - auc: 0.9919 - recall: 0.9983 - val_loss: 0.2940 - val_acc: 0.9157 - val_auc: 0.9920 - val_recall: 0.9984\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00021: early stopping\n",
      "training time: 157.6658275127411s\n",
      "Train on 5881 samples, validate on 1471 samples\n",
      "Epoch 1/60\n",
      "5881/5881 [==============================] - 9s 1ms/step - loss: 1.3190 - acc: 0.6251 - auc: 0.8018 - recall: 0.9562 - val_loss: 0.7225 - val_acc: 0.7254 - val_auc: 0.9041 - val_recall: 0.9865\n",
      "Epoch 2/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.5088 - acc: 0.8050 - auc: 0.9275 - recall: 0.9912 - val_loss: 0.5307 - val_acc: 0.8185 - val_auc: 0.9445 - val_recall: 0.9937\n",
      "Epoch 3/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.3713 - acc: 0.8677 - auc: 0.9535 - recall: 0.9949 - val_loss: 0.4618 - val_acc: 0.8396 - val_auc: 0.9603 - val_recall: 0.9953\n",
      "Epoch 4/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.2698 - acc: 0.8992 - auc: 0.9652 - recall: 0.9960 - val_loss: 0.3973 - val_acc: 0.8600 - val_auc: 0.9692 - val_recall: 0.9965\n",
      "Epoch 5/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.2237 - acc: 0.9129 - auc: 0.9724 - recall: 0.9969 - val_loss: 0.3682 - val_acc: 0.8770 - val_auc: 0.9750 - val_recall: 0.9972\n",
      "Epoch 6/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1957 - acc: 0.9288 - auc: 0.9771 - recall: 0.9975 - val_loss: 0.3525 - val_acc: 0.8919 - val_auc: 0.9789 - val_recall: 0.9977\n",
      "Epoch 7/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1781 - acc: 0.9373 - auc: 0.9803 - recall: 0.9979 - val_loss: 0.3450 - val_acc: 0.8939 - val_auc: 0.9816 - val_recall: 0.9980\n",
      "Epoch 8/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1636 - acc: 0.9390 - auc: 0.9827 - recall: 0.9981 - val_loss: 0.3427 - val_acc: 0.8919 - val_auc: 0.9836 - val_recall: 0.9983\n",
      "\n",
      "Epoch 00008: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 9/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1554 - acc: 0.9435 - auc: 0.9844 - recall: 0.9984 - val_loss: 0.3317 - val_acc: 0.8980 - val_auc: 0.9851 - val_recall: 0.9985\n",
      "Epoch 10/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.2159 - acc: 0.9447 - auc: 0.9857 - recall: 0.9984 - val_loss: 0.3369 - val_acc: 0.8919 - val_auc: 0.9861 - val_recall: 0.9982\n",
      "Epoch 11/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.2806 - acc: 0.9415 - auc: 0.9864 - recall: 0.9980 - val_loss: 0.3565 - val_acc: 0.8912 - val_auc: 0.9866 - val_recall: 0.9977\n",
      "Epoch 12/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.2908 - acc: 0.9410 - auc: 0.9868 - recall: 0.9975 - val_loss: 0.4250 - val_acc: 0.8906 - val_auc: 0.9869 - val_recall: 0.9972\n",
      "\n",
      "Epoch 00012: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 13/60\n",
      "5881/5881 [==============================] - 6s 1ms/step - loss: 0.2363 - acc: 0.9442 - auc: 0.9870 - recall: 0.9970 - val_loss: 0.3298 - val_acc: 0.8892 - val_auc: 0.9873 - val_recall: 0.9969\n",
      "Epoch 14/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1319 - acc: 0.9509 - auc: 0.9877 - recall: 0.9970 - val_loss: 0.3199 - val_acc: 0.8967 - val_auc: 0.9880 - val_recall: 0.9971\n",
      "Epoch 15/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1261 - acc: 0.9551 - auc: 0.9883 - recall: 0.9972 - val_loss: 0.3195 - val_acc: 0.8946 - val_auc: 0.9886 - val_recall: 0.9973\n",
      "\n",
      "Epoch 00015: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 16/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1266 - acc: 0.9524 - auc: 0.9889 - recall: 0.9974 - val_loss: 0.3219 - val_acc: 0.8987 - val_auc: 0.9891 - val_recall: 0.9975\n",
      "Epoch 17/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1228 - acc: 0.9551 - auc: 0.9893 - recall: 0.9976 - val_loss: 0.3208 - val_acc: 0.8994 - val_auc: 0.9896 - val_recall: 0.9976\n",
      "Epoch 18/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1209 - acc: 0.9551 - auc: 0.9898 - recall: 0.9977 - val_loss: 0.3193 - val_acc: 0.8967 - val_auc: 0.9900 - val_recall: 0.9977\n",
      "\n",
      "Epoch 00018: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\n",
      "Epoch 19/60\n",
      "5881/5881 [==============================] - 7s 1ms/step - loss: 0.1216 - acc: 0.9558 - auc: 0.9901 - recall: 0.9978 - val_loss: 0.3190 - val_acc: 0.8973 - val_auc: 0.9903 - val_recall: 0.9979\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00019: early stopping\n",
      "training time: 145.94875741004944s\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_17 (InputLayer)        (None, 128, 9)            0         \n",
      "_________________________________________________________________\n",
      "dense_122 (Dense)            (None, 128, 250)          2500      \n",
      "_________________________________________________________________\n",
      "dropout_23 (Dropout)         (None, 128, 250)          0         \n",
      "_________________________________________________________________\n",
      "dense_123 (Dense)            (None, 128, 250)          62750     \n",
      "_________________________________________________________________\n",
      "dropout_24 (Dropout)         (None, 128, 250)          0         \n",
      "_________________________________________________________________\n",
      "dense_124 (Dense)            (None, 128, 30)           7530      \n",
      "_________________________________________________________________\n",
      "flatten_10 (Flatten)         (None, 3840)              0         \n",
      "_________________________________________________________________\n",
      "dense_125 (Dense)            (None, 6)                 23046     \n",
      "=================================================================\n",
      "Total params: 95,826\n",
      "Trainable params: 95,826\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Train on 5881 samples, validate on 1471 samples\n",
      "Epoch 1/60\n",
      "5881/5881 [==============================] - 5s 809us/step - loss: 1.1009 - acc: 0.5530 - auc: 0.7329 - recall: 0.9575 - val_loss: 1.0442 - val_acc: 0.6893 - val_auc: 0.8902 - val_recall: 1.0000\n",
      "Epoch 2/60\n",
      "5881/5881 [==============================] - 3s 584us/step - loss: 0.5196 - acc: 0.7723 - auc: 0.9182 - recall: 1.0000 - val_loss: 0.8964 - val_acc: 0.7641 - val_auc: 0.9352 - val_recall: 1.0000\n",
      "Epoch 3/60\n",
      "5881/5881 [==============================] - 3s 572us/step - loss: 0.4021 - acc: 0.8390 - auc: 0.9451 - recall: 1.0000 - val_loss: 0.7785 - val_acc: 0.8382 - val_auc: 0.9529 - val_recall: 1.0000\n",
      "Epoch 4/60\n",
      "5881/5881 [==============================] - 3s 572us/step - loss: 0.3316 - acc: 0.8721 - auc: 0.9585 - recall: 1.0000 - val_loss: 0.7293 - val_acc: 0.8742 - val_auc: 0.9628 - val_recall: 1.0000\n",
      "Epoch 5/60\n",
      "5881/5881 [==============================] - 3s 571us/step - loss: 0.2809 - acc: 0.8930 - auc: 0.9662 - recall: 1.0000 - val_loss: 0.7018 - val_acc: 0.8715 - val_auc: 0.9690 - val_recall: 1.0000\n",
      "Epoch 6/60\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5881/5881 [==============================] - 3s 568us/step - loss: 0.2351 - acc: 0.9155 - auc: 0.9713 - recall: 1.0000 - val_loss: 0.6875 - val_acc: 0.8838 - val_auc: 0.9734 - val_recall: 1.0000\n",
      "Epoch 7/60\n",
      "5881/5881 [==============================] - 3s 572us/step - loss: 0.1891 - acc: 0.9337 - auc: 0.9752 - recall: 1.0000 - val_loss: 0.7050 - val_acc: 0.8749 - val_auc: 0.9767 - val_recall: 1.0000\n",
      "Epoch 8/60\n",
      "5881/5881 [==============================] - 3s 575us/step - loss: 0.1890 - acc: 0.9281 - auc: 0.9778 - recall: 1.0000 - val_loss: 0.6712 - val_acc: 0.8939 - val_auc: 0.9789 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00008: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 9/60\n",
      "5881/5881 [==============================] - 3s 569us/step - loss: 0.1502 - acc: 0.9434 - auc: 0.9799 - recall: 1.0000 - val_loss: 0.6574 - val_acc: 0.9069 - val_auc: 0.9809 - val_recall: 1.0000\n",
      "Epoch 10/60\n",
      "5881/5881 [==============================] - 3s 571us/step - loss: 0.1367 - acc: 0.9449 - auc: 0.9817 - recall: 1.0000 - val_loss: 0.6578 - val_acc: 0.9089 - val_auc: 0.9824 - val_recall: 1.0000\n",
      "Epoch 11/60\n",
      "5881/5881 [==============================] - 3s 568us/step - loss: 0.1304 - acc: 0.9498 - auc: 0.9831 - recall: 1.0000 - val_loss: 0.6675 - val_acc: 0.8987 - val_auc: 0.9837 - val_recall: 1.0000\n",
      "Epoch 12/60\n",
      "5881/5881 [==============================] - 3s 572us/step - loss: 0.1263 - acc: 0.9495 - auc: 0.9842 - recall: 1.0000 - val_loss: 0.6583 - val_acc: 0.9089 - val_auc: 0.9847 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00012: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 13/60\n",
      "5881/5881 [==============================] - 3s 569us/step - loss: 0.1221 - acc: 0.9481 - auc: 0.9852 - recall: 1.0000 - val_loss: 0.6574 - val_acc: 0.9075 - val_auc: 0.9855 - val_recall: 1.0000\n",
      "Epoch 14/60\n",
      "5881/5881 [==============================] - 3s 570us/step - loss: 0.1150 - acc: 0.9546 - auc: 0.9859 - recall: 1.0000 - val_loss: 0.6667 - val_acc: 0.9069 - val_auc: 0.9862 - val_recall: 1.0000\n",
      "Epoch 15/60\n",
      "5881/5881 [==============================] - 3s 562us/step - loss: 0.1168 - acc: 0.9519 - auc: 0.9865 - recall: 1.0000 - val_loss: 0.6599 - val_acc: 0.9075 - val_auc: 0.9868 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00015: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 16/60\n",
      "5881/5881 [==============================] - 3s 572us/step - loss: 0.1105 - acc: 0.9543 - auc: 0.9871 - recall: 1.0000 - val_loss: 0.6591 - val_acc: 0.9082 - val_auc: 0.9873 - val_recall: 1.0000\n",
      "Epoch 17/60\n",
      "5881/5881 [==============================] - 3s 568us/step - loss: 0.1092 - acc: 0.9553 - auc: 0.9876 - recall: 1.0000 - val_loss: 0.6663 - val_acc: 0.9048 - val_auc: 0.9877 - val_recall: 1.0000\n",
      "Epoch 18/60\n",
      "5881/5881 [==============================] - 3s 571us/step - loss: 0.1080 - acc: 0.9554 - auc: 0.9879 - recall: 1.0000 - val_loss: 0.6564 - val_acc: 0.9096 - val_auc: 0.9881 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00018: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\n",
      "Epoch 19/60\n",
      "5881/5881 [==============================] - 3s 567us/step - loss: 0.1076 - acc: 0.9548 - auc: 0.9883 - recall: 1.0000 - val_loss: 0.6582 - val_acc: 0.9089 - val_auc: 0.9884 - val_recall: 1.0000\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00019: early stopping\n",
      "training time: 84.30107402801514s\n",
      "Train on 5881 samples, validate on 1471 samples\n",
      "Epoch 1/60\n",
      "5881/5881 [==============================] - 5s 790us/step - loss: 1.0907 - acc: 0.5786 - auc: 0.7418 - recall: 0.9575 - val_loss: 0.8389 - val_acc: 0.7498 - val_auc: 0.8973 - val_recall: 1.0000\n",
      "Epoch 2/60\n",
      "5881/5881 [==============================] - 3s 571us/step - loss: 0.5133 - acc: 0.7735 - auc: 0.9232 - recall: 1.0000 - val_loss: 0.7191 - val_acc: 0.7634 - val_auc: 0.9383 - val_recall: 1.0000\n",
      "Epoch 3/60\n",
      "5881/5881 [==============================] - 3s 563us/step - loss: 0.4139 - acc: 0.8267 - auc: 0.9463 - recall: 1.0000 - val_loss: 0.5916 - val_acc: 0.8443 - val_auc: 0.9538 - val_recall: 1.0000\n",
      "Epoch 4/60\n",
      "5881/5881 [==============================] - 3s 571us/step - loss: 0.3385 - acc: 0.8708 - auc: 0.9589 - recall: 1.0000 - val_loss: 0.5904 - val_acc: 0.8498 - val_auc: 0.9630 - val_recall: 1.0000\n",
      "Epoch 5/60\n",
      "5881/5881 [==============================] - 3s 566us/step - loss: 0.2828 - acc: 0.8925 - auc: 0.9661 - recall: 1.0000 - val_loss: 0.5471 - val_acc: 0.8872 - val_auc: 0.9690 - val_recall: 1.0000\n",
      "Epoch 6/60\n",
      "5881/5881 [==============================] - 3s 559us/step - loss: 0.2333 - acc: 0.9163 - auc: 0.9714 - recall: 1.0000 - val_loss: 0.6113 - val_acc: 0.8545 - val_auc: 0.9734 - val_recall: 1.0000\n",
      "Epoch 7/60\n",
      "5881/5881 [==============================] - 3s 560us/step - loss: 0.2115 - acc: 0.9153 - auc: 0.9749 - recall: 1.0000 - val_loss: 0.5616 - val_acc: 0.8858 - val_auc: 0.9764 - val_recall: 1.0000\n",
      "Epoch 8/60\n",
      "5881/5881 [==============================] - 3s 568us/step - loss: 0.1724 - acc: 0.9344 - auc: 0.9778 - recall: 1.0000 - val_loss: 0.5704 - val_acc: 0.8967 - val_auc: 0.9789 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00008: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 9/60\n",
      "5881/5881 [==============================] - 3s 569us/step - loss: 0.1537 - acc: 0.9437 - auc: 0.9800 - recall: 1.0000 - val_loss: 0.6190 - val_acc: 0.8987 - val_auc: 0.9809 - val_recall: 1.0000\n",
      "Epoch 10/60\n",
      "5881/5881 [==============================] - 3s 566us/step - loss: 0.1418 - acc: 0.9485 - auc: 0.9818 - recall: 1.0000 - val_loss: 0.6251 - val_acc: 0.8973 - val_auc: 0.9825 - val_recall: 1.0000\n",
      "Epoch 11/60\n",
      "5881/5881 [==============================] - 3s 565us/step - loss: 0.1355 - acc: 0.9466 - auc: 0.9831 - recall: 1.0000 - val_loss: 0.6386 - val_acc: 0.9055 - val_auc: 0.9837 - val_recall: 1.0000\n",
      "Epoch 12/60\n",
      "5881/5881 [==============================] - 3s 566us/step - loss: 0.1248 - acc: 0.9522 - auc: 0.9843 - recall: 1.0000 - val_loss: 0.6440 - val_acc: 0.8980 - val_auc: 0.9847 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00012: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 13/60\n",
      "5881/5881 [==============================] - 3s 566us/step - loss: 0.1208 - acc: 0.9544 - auc: 0.9851 - recall: 1.0000 - val_loss: 0.6592 - val_acc: 0.9041 - val_auc: 0.9855 - val_recall: 1.0000\n",
      "Epoch 14/60\n",
      "5881/5881 [==============================] - 3s 569us/step - loss: 0.1138 - acc: 0.9568 - auc: 0.9859 - recall: 1.0000 - val_loss: 0.6565 - val_acc: 0.9082 - val_auc: 0.9862 - val_recall: 1.0000\n",
      "Epoch 15/60\n",
      "5881/5881 [==============================] - 3s 567us/step - loss: 0.1118 - acc: 0.9575 - auc: 0.9865 - recall: 1.0000 - val_loss: 0.6584 - val_acc: 0.9028 - val_auc: 0.9868 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00015: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 16/60\n",
      "5881/5881 [==============================] - 3s 566us/step - loss: 0.1119 - acc: 0.9548 - auc: 0.9870 - recall: 1.0000 - val_loss: 0.6705 - val_acc: 0.9075 - val_auc: 0.9873 - val_recall: 1.0000\n",
      "Epoch 17/60\n",
      "5881/5881 [==============================] - 3s 569us/step - loss: 0.1076 - acc: 0.9548 - auc: 0.9875 - recall: 1.0000 - val_loss: 0.6738 - val_acc: 0.9069 - val_auc: 0.9877 - val_recall: 1.0000\n",
      "Epoch 18/60\n",
      "5881/5881 [==============================] - 3s 569us/step - loss: 0.1070 - acc: 0.9592 - auc: 0.9879 - recall: 1.0000 - val_loss: 0.6751 - val_acc: 0.9082 - val_auc: 0.9881 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00018: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\n",
      "Epoch 19/60\n",
      "5881/5881 [==============================] - 3s 568us/step - loss: 0.1039 - acc: 0.9583 - auc: 0.9883 - recall: 1.0000 - val_loss: 0.6726 - val_acc: 0.9075 - val_auc: 0.9884 - val_recall: 1.0000\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00019: early stopping\n",
      "training time: 83.7048978805542s\n",
      "Train on 5881 samples, validate on 1471 samples\n",
      "Epoch 1/60\n",
      "5881/5881 [==============================] - 5s 787us/step - loss: 1.1639 - acc: 0.5242 - auc: 0.7073 - recall: 0.9575 - val_loss: 0.8068 - val_acc: 0.7104 - val_auc: 0.8765 - val_recall: 1.0000\n",
      "Epoch 2/60\n",
      "5881/5881 [==============================] - 3s 561us/step - loss: 0.5071 - acc: 0.7781 - auc: 0.9126 - recall: 1.0000 - val_loss: 0.7209 - val_acc: 0.7356 - val_auc: 0.9327 - val_recall: 1.0000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 3/60\n",
      "5881/5881 [==============================] - 3s 564us/step - loss: 0.3983 - acc: 0.8417 - auc: 0.9429 - recall: 1.0000 - val_loss: 0.6373 - val_acc: 0.8389 - val_auc: 0.9515 - val_recall: 1.0000\n",
      "Epoch 4/60\n",
      "5881/5881 [==============================] - 3s 565us/step - loss: 0.3223 - acc: 0.8777 - auc: 0.9573 - recall: 1.0000 - val_loss: 0.6238 - val_acc: 0.8647 - val_auc: 0.9620 - val_recall: 1.0000\n",
      "Epoch 5/60\n",
      "5881/5881 [==============================] - 3s 563us/step - loss: 0.2777 - acc: 0.8953 - auc: 0.9654 - recall: 1.0000 - val_loss: 0.6893 - val_acc: 0.8688 - val_auc: 0.9682 - val_recall: 1.0000\n",
      "Epoch 6/60\n",
      "5881/5881 [==============================] - 3s 563us/step - loss: 0.2332 - acc: 0.9114 - auc: 0.9706 - recall: 1.0000 - val_loss: 0.6565 - val_acc: 0.8960 - val_auc: 0.9727 - val_recall: 1.0000\n",
      "Epoch 7/60\n",
      "5881/5881 [==============================] - 3s 564us/step - loss: 0.1783 - acc: 0.9376 - auc: 0.9747 - recall: 1.0000 - val_loss: 0.6742 - val_acc: 0.9041 - val_auc: 0.9763 - val_recall: 1.0000\n",
      "Epoch 8/60\n",
      "5881/5881 [==============================] - 3s 565us/step - loss: 0.1622 - acc: 0.9359 - auc: 0.9776 - recall: 1.0000 - val_loss: 0.6845 - val_acc: 0.8994 - val_auc: 0.9788 - val_recall: 1.0000\n",
      "Epoch 9/60\n",
      "5881/5881 [==============================] - 3s 565us/step - loss: 0.1532 - acc: 0.9393 - auc: 0.9798 - recall: 1.0000 - val_loss: 0.6820 - val_acc: 0.9001 - val_auc: 0.9807 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00009: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 10/60\n",
      "5881/5881 [==============================] - 3s 565us/step - loss: 0.1250 - acc: 0.9515 - auc: 0.9816 - recall: 1.0000 - val_loss: 0.6715 - val_acc: 0.9082 - val_auc: 0.9823 - val_recall: 1.0000\n",
      "Epoch 11/60\n",
      "5881/5881 [==============================] - 3s 560us/step - loss: 0.1145 - acc: 0.9558 - auc: 0.9829 - recall: 1.0000 - val_loss: 0.6683 - val_acc: 0.9041 - val_auc: 0.9835 - val_recall: 1.0000\n",
      "Epoch 12/60\n",
      "5881/5881 [==============================] - 3s 574us/step - loss: 0.1120 - acc: 0.9582 - auc: 0.9841 - recall: 1.0000 - val_loss: 0.6621 - val_acc: 0.9089 - val_auc: 0.9845 - val_recall: 1.0000\n",
      "Epoch 13/60\n",
      "5881/5881 [==============================] - 3s 578us/step - loss: 0.1079 - acc: 0.9568 - auc: 0.9850 - recall: 1.0000 - val_loss: 0.6669 - val_acc: 0.9055 - val_auc: 0.9853 - val_recall: 1.0000\n",
      "Epoch 14/60\n",
      "5881/5881 [==============================] - 3s 562us/step - loss: 0.1098 - acc: 0.9578 - auc: 0.9857 - recall: 1.0000 - val_loss: 0.6788 - val_acc: 0.9082 - val_auc: 0.9860 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00014: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 15/60\n",
      "5881/5881 [==============================] - 3s 559us/step - loss: 0.1006 - acc: 0.9597 - auc: 0.9863 - recall: 1.0000 - val_loss: 0.6769 - val_acc: 0.9062 - val_auc: 0.9866 - val_recall: 1.0000\n",
      "Epoch 16/60\n",
      "5881/5881 [==============================] - 3s 564us/step - loss: 0.0985 - acc: 0.9600 - auc: 0.9869 - recall: 1.0000 - val_loss: 0.6736 - val_acc: 0.9048 - val_auc: 0.9871 - val_recall: 1.0000\n",
      "Epoch 17/60\n",
      "5881/5881 [==============================] - 3s 563us/step - loss: 0.0966 - acc: 0.9626 - auc: 0.9873 - recall: 1.0000 - val_loss: 0.6822 - val_acc: 0.9089 - val_auc: 0.9875 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00017: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 18/60\n",
      "5881/5881 [==============================] - 3s 561us/step - loss: 0.0957 - acc: 0.9617 - auc: 0.9877 - recall: 1.0000 - val_loss: 0.6753 - val_acc: 0.9089 - val_auc: 0.9878 - val_recall: 1.0000\n",
      "Epoch 19/60\n",
      "5881/5881 [==============================] - 3s 561us/step - loss: 0.0956 - acc: 0.9619 - auc: 0.9880 - recall: 1.0000 - val_loss: 0.6838 - val_acc: 0.9089 - val_auc: 0.9882 - val_recall: 1.0000\n",
      "Epoch 20/60\n",
      "5881/5881 [==============================] - 3s 558us/step - loss: 0.0949 - acc: 0.9609 - auc: 0.9883 - recall: 1.0000 - val_loss: 0.6858 - val_acc: 0.9089 - val_auc: 0.9884 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00020: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\n",
      "Epoch 21/60\n",
      "5881/5881 [==============================] - 3s 566us/step - loss: 0.0928 - acc: 0.9629 - auc: 0.9886 - recall: 1.0000 - val_loss: 0.6777 - val_acc: 0.9075 - val_auc: 0.9887 - val_recall: 1.0000\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00021: early stopping\n",
      "training time: 90.58042311668396s\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_20 (InputLayer)        (None, 128, 9)            0         \n",
      "_________________________________________________________________\n",
      "conv1d_1 (Conv1D)            (None, 119, 100)          9100      \n",
      "_________________________________________________________________\n",
      "conv1d_2 (Conv1D)            (None, 110, 100)          100100    \n",
      "_________________________________________________________________\n",
      "max_pooling1d_1 (MaxPooling1 (None, 36, 100)           0         \n",
      "_________________________________________________________________\n",
      "conv1d_3 (Conv1D)            (None, 27, 160)           160160    \n",
      "_________________________________________________________________\n",
      "max_pooling1d_2 (MaxPooling1 (None, 9, 160)            0         \n",
      "_________________________________________________________________\n",
      "flatten_13 (Flatten)         (None, 1440)              0         \n",
      "_________________________________________________________________\n",
      "dense_134 (Dense)            (None, 20)                28820     \n",
      "_________________________________________________________________\n",
      "dense_135 (Dense)            (None, 6)                 126       \n",
      "=================================================================\n",
      "Total params: 298,306\n",
      "Trainable params: 298,306\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Train on 5881 samples, validate on 1471 samples\n",
      "Epoch 1/60\n",
      "5881/5881 [==============================] - 4s 620us/step - loss: 1.1523 - acc: 0.4423 - auc: 0.7283 - recall: 0.9575 - val_loss: 0.8609 - val_acc: 0.6241 - val_auc: 0.8670 - val_recall: 1.0000\n",
      "Epoch 2/60\n",
      "5881/5881 [==============================] - 2s 402us/step - loss: 0.4542 - acc: 0.8397 - auc: 0.9071 - recall: 1.0000 - val_loss: 0.4491 - val_acc: 0.8668 - val_auc: 0.9368 - val_recall: 1.0000\n",
      "Epoch 3/60\n",
      "5881/5881 [==============================] - 2s 399us/step - loss: 0.1928 - acc: 0.9310 - auc: 0.9529 - recall: 1.0000 - val_loss: 0.3029 - val_acc: 0.9109 - val_auc: 0.9641 - val_recall: 1.0000\n",
      "Epoch 4/60\n",
      "5881/5881 [==============================] - 2s 402us/step - loss: 0.1236 - acc: 0.9522 - auc: 0.9710 - recall: 1.0000 - val_loss: 0.3661 - val_acc: 0.9021 - val_auc: 0.9758 - val_recall: 1.0000\n",
      "Epoch 5/60\n",
      "5881/5881 [==============================] - 2s 396us/step - loss: 0.1078 - acc: 0.9570 - auc: 0.9792 - recall: 1.0000 - val_loss: 0.3195 - val_acc: 0.9082 - val_auc: 0.9816 - val_recall: 1.0000\n",
      "Epoch 6/60\n",
      "5881/5881 [==============================] - 2s 398us/step - loss: 0.1037 - acc: 0.9544 - auc: 0.9835 - recall: 1.0000 - val_loss: 0.3255 - val_acc: 0.9164 - val_auc: 0.9850 - val_recall: 1.0000\n",
      "Epoch 7/60\n",
      "5881/5881 [==============================] - 2s 399us/step - loss: 0.1175 - acc: 0.9561 - auc: 0.9861 - recall: 1.0000 - val_loss: 0.3415 - val_acc: 0.9137 - val_auc: 0.9870 - val_recall: 1.0000\n",
      "Epoch 8/60\n",
      "5881/5881 [==============================] - 2s 395us/step - loss: 0.0959 - acc: 0.9590 - auc: 0.9879 - recall: 1.0000 - val_loss: 0.3244 - val_acc: 0.9273 - val_auc: 0.9885 - val_recall: 1.0000\n",
      "Epoch 9/60\n",
      "5881/5881 [==============================] - 2s 397us/step - loss: 0.0887 - acc: 0.9636 - auc: 0.9892 - recall: 1.0000 - val_loss: 0.3255 - val_acc: 0.9300 - val_auc: 0.9896 - val_recall: 1.0000\n",
      "Epoch 10/60\n",
      "5881/5881 [==============================] - 2s 394us/step - loss: 0.0911 - acc: 0.9583 - auc: 0.9900 - recall: 1.0000 - val_loss: 0.3522 - val_acc: 0.9198 - val_auc: 0.9903 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00010: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 11/60\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5881/5881 [==============================] - 2s 397us/step - loss: 0.0853 - acc: 0.9609 - auc: 0.9907 - recall: 1.0000 - val_loss: 0.3462 - val_acc: 0.9205 - val_auc: 0.9909 - val_recall: 1.0000\n",
      "Epoch 12/60\n",
      "5881/5881 [==============================] - 2s 398us/step - loss: 0.0866 - acc: 0.9600 - auc: 0.9912 - recall: 1.0000 - val_loss: 0.3406 - val_acc: 0.9279 - val_auc: 0.9914 - val_recall: 1.0000\n",
      "Epoch 13/60\n",
      "5881/5881 [==============================] - 2s 397us/step - loss: 0.0822 - acc: 0.9628 - auc: 0.9916 - recall: 1.0000 - val_loss: 0.3396 - val_acc: 0.9225 - val_auc: 0.9918 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00013: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 14/60\n",
      "5881/5881 [==============================] - 2s 396us/step - loss: 0.0812 - acc: 0.9633 - auc: 0.9920 - recall: 1.0000 - val_loss: 0.3580 - val_acc: 0.9205 - val_auc: 0.9921 - val_recall: 1.0000\n",
      "Epoch 15/60\n",
      "5881/5881 [==============================] - 2s 401us/step - loss: 0.0813 - acc: 0.9634 - auc: 0.9922 - recall: 1.0000 - val_loss: 0.3559 - val_acc: 0.9130 - val_auc: 0.9924 - val_recall: 1.0000\n",
      "Epoch 16/60\n",
      "5881/5881 [==============================] - 2s 394us/step - loss: 0.0808 - acc: 0.9629 - auc: 0.9925 - recall: 1.0000 - val_loss: 0.3695 - val_acc: 0.9171 - val_auc: 0.9926 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00016: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 17/60\n",
      "5881/5881 [==============================] - 2s 399us/step - loss: 0.0798 - acc: 0.9617 - auc: 0.9927 - recall: 1.0000 - val_loss: 0.3701 - val_acc: 0.9157 - val_auc: 0.9927 - val_recall: 1.0000\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00017: early stopping\n",
      "training time: 61.17898488044739s\n",
      "Train on 5881 samples, validate on 1471 samples\n",
      "Epoch 1/60\n",
      "5881/5881 [==============================] - 4s 619us/step - loss: 0.9808 - acc: 0.5922 - auc: 0.7492 - recall: 0.9575 - val_loss: 0.8925 - val_acc: 0.6873 - val_auc: 0.9073 - val_recall: 1.0000\n",
      "Epoch 2/60\n",
      "5881/5881 [==============================] - 2s 399us/step - loss: 0.2967 - acc: 0.8881 - auc: 0.9343 - recall: 1.0000 - val_loss: 0.2681 - val_acc: 0.9137 - val_auc: 0.9577 - val_recall: 1.0000\n",
      "Epoch 3/60\n",
      "5881/5881 [==============================] - 2s 396us/step - loss: 0.1415 - acc: 0.9434 - auc: 0.9690 - recall: 1.0000 - val_loss: 0.2790 - val_acc: 0.9082 - val_auc: 0.9761 - val_recall: 1.0000\n",
      "Epoch 4/60\n",
      "5881/5881 [==============================] - 2s 396us/step - loss: 0.1211 - acc: 0.9492 - auc: 0.9803 - recall: 1.0000 - val_loss: 0.2697 - val_acc: 0.9239 - val_auc: 0.9835 - val_recall: 1.0000\n",
      "Epoch 5/60\n",
      "5881/5881 [==============================] - 2s 393us/step - loss: 0.0995 - acc: 0.9589 - auc: 0.9857 - recall: 1.0000 - val_loss: 0.2718 - val_acc: 0.9191 - val_auc: 0.9874 - val_recall: 1.0000\n",
      "Epoch 6/60\n",
      "5881/5881 [==============================] - 2s 397us/step - loss: 0.0903 - acc: 0.9602 - auc: 0.9887 - recall: 1.0000 - val_loss: 0.2529 - val_acc: 0.9245 - val_auc: 0.9897 - val_recall: 1.0000\n",
      "Epoch 7/60\n",
      "5881/5881 [==============================] - 2s 393us/step - loss: 0.0899 - acc: 0.9578 - auc: 0.9906 - recall: 1.0000 - val_loss: 0.2721 - val_acc: 0.9232 - val_auc: 0.9912 - val_recall: 1.0000\n",
      "Epoch 8/60\n",
      "5881/5881 [==============================] - 2s 396us/step - loss: 0.0965 - acc: 0.9549 - auc: 0.9918 - recall: 1.0000 - val_loss: 0.2882 - val_acc: 0.8973 - val_auc: 0.9922 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00008: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 9/60\n",
      "5881/5881 [==============================] - 2s 394us/step - loss: 0.0976 - acc: 0.9549 - auc: 0.9927 - recall: 1.0000 - val_loss: 0.2994 - val_acc: 0.9143 - val_auc: 0.9929 - val_recall: 1.0000\n",
      "Epoch 10/60\n",
      "5881/5881 [==============================] - 2s 401us/step - loss: 0.0869 - acc: 0.9583 - auc: 0.9932 - recall: 1.0000 - val_loss: 0.3154 - val_acc: 0.9082 - val_auc: 0.9933 - val_recall: 1.0000\n",
      "Epoch 11/60\n",
      "5881/5881 [==============================] - 2s 393us/step - loss: 0.0890 - acc: 0.9577 - auc: 0.9934 - recall: 1.0000 - val_loss: 0.3032 - val_acc: 0.9130 - val_auc: 0.9936 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00011: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 12/60\n",
      "5881/5881 [==============================] - 2s 398us/step - loss: 0.0815 - acc: 0.9617 - auc: 0.9938 - recall: 1.0000 - val_loss: 0.3004 - val_acc: 0.9082 - val_auc: 0.9938 - val_recall: 1.0000\n",
      "Epoch 13/60\n",
      "5881/5881 [==============================] - 2s 397us/step - loss: 0.0784 - acc: 0.9624 - auc: 0.9940 - recall: 1.0000 - val_loss: 0.3089 - val_acc: 0.9130 - val_auc: 0.9940 - val_recall: 1.0000\n",
      "Epoch 14/60\n",
      "5881/5881 [==============================] - 2s 398us/step - loss: 0.0775 - acc: 0.9619 - auc: 0.9942 - recall: 1.0000 - val_loss: 0.3033 - val_acc: 0.9157 - val_auc: 0.9942 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00014: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 15/60\n",
      "5881/5881 [==============================] - 2s 399us/step - loss: 0.0761 - acc: 0.9650 - auc: 0.9943 - recall: 1.0000 - val_loss: 0.3115 - val_acc: 0.9109 - val_auc: 0.9944 - val_recall: 1.0000\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00015: early stopping\n",
      "training time: 56.98405480384827s\n",
      "Train on 5881 samples, validate on 1471 samples\n",
      "Epoch 1/60\n",
      "5881/5881 [==============================] - 4s 641us/step - loss: 0.8919 - acc: 0.6041 - auc: 0.7615 - recall: 0.9575 - val_loss: 0.7076 - val_acc: 0.8137 - val_auc: 0.9243 - val_recall: 1.0000\n",
      "Epoch 2/60\n",
      "5881/5881 [==============================] - 2s 400us/step - loss: 0.2846 - acc: 0.8871 - auc: 0.9506 - recall: 1.0000 - val_loss: 0.4585 - val_acc: 0.9014 - val_auc: 0.9656 - val_recall: 1.0000\n",
      "Epoch 3/60\n",
      "5881/5881 [==============================] - 2s 407us/step - loss: 0.1567 - acc: 0.9390 - auc: 0.9733 - recall: 1.0000 - val_loss: 0.3605 - val_acc: 0.9055 - val_auc: 0.9786 - val_recall: 1.0000\n",
      "Epoch 4/60\n",
      "5881/5881 [==============================] - 2s 401us/step - loss: 0.1097 - acc: 0.9578 - auc: 0.9821 - recall: 1.0000 - val_loss: 0.3179 - val_acc: 0.9184 - val_auc: 0.9846 - val_recall: 1.0000\n",
      "Epoch 5/60\n",
      "5881/5881 [==============================] - 2s 403us/step - loss: 0.1026 - acc: 0.9532 - auc: 0.9864 - recall: 1.0000 - val_loss: 0.3460 - val_acc: 0.9164 - val_auc: 0.9877 - val_recall: 1.0000\n",
      "Epoch 6/60\n",
      "5881/5881 [==============================] - 2s 402us/step - loss: 0.0906 - acc: 0.9597 - auc: 0.9888 - recall: 1.0000 - val_loss: 0.3914 - val_acc: 0.9143 - val_auc: 0.9896 - val_recall: 1.0000\n",
      "Epoch 7/60\n",
      "5881/5881 [==============================] - 2s 399us/step - loss: 0.0885 - acc: 0.9606 - auc: 0.9902 - recall: 1.0000 - val_loss: 0.3570 - val_acc: 0.9293 - val_auc: 0.9906 - val_recall: 1.0000\n",
      "Epoch 8/60\n",
      "5881/5881 [==============================] - 2s 405us/step - loss: 0.0872 - acc: 0.9580 - auc: 0.9911 - recall: 1.0000 - val_loss: 0.3948 - val_acc: 0.9300 - val_auc: 0.9914 - val_recall: 1.0000\n",
      "Epoch 9/60\n",
      "5881/5881 [==============================] - 2s 399us/step - loss: 0.0846 - acc: 0.9585 - auc: 0.9917 - recall: 1.0000 - val_loss: 0.4528 - val_acc: 0.9293 - val_auc: 0.9919 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00009: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 10/60\n",
      "5881/5881 [==============================] - 2s 403us/step - loss: 0.0782 - acc: 0.9617 - auc: 0.9922 - recall: 1.0000 - val_loss: 0.4578 - val_acc: 0.9279 - val_auc: 0.9923 - val_recall: 1.0000\n",
      "Epoch 11/60\n",
      "5881/5881 [==============================] - 2s 406us/step - loss: 0.0747 - acc: 0.9621 - auc: 0.9925 - recall: 1.0000 - val_loss: 0.4496 - val_acc: 0.9293 - val_auc: 0.9926 - val_recall: 1.0000\n",
      "Epoch 12/60\n",
      "5881/5881 [==============================] - 2s 400us/step - loss: 0.0714 - acc: 0.9638 - auc: 0.9928 - recall: 1.0000 - val_loss: 0.4820 - val_acc: 0.9191 - val_auc: 0.9929 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00012: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 13/60\n",
      "5881/5881 [==============================] - 2s 406us/step - loss: 0.0706 - acc: 0.9636 - auc: 0.9930 - recall: 1.0000 - val_loss: 0.4874 - val_acc: 0.9211 - val_auc: 0.9931 - val_recall: 1.0000\n",
      "Epoch 14/60\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5881/5881 [==============================] - 2s 401us/step - loss: 0.0680 - acc: 0.9646 - auc: 0.9932 - recall: 1.0000 - val_loss: 0.4969 - val_acc: 0.9191 - val_auc: 0.9932 - val_recall: 1.0000\n",
      "Epoch 15/60\n",
      "5881/5881 [==============================] - 2s 401us/step - loss: 0.0682 - acc: 0.9650 - auc: 0.9933 - recall: 1.0000 - val_loss: 0.5054 - val_acc: 0.9137 - val_auc: 0.9934 - val_recall: 1.0000\n",
      "\n",
      "Epoch 00015: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 16/60\n",
      "5881/5881 [==============================] - 2s 405us/step - loss: 0.0665 - acc: 0.9662 - auc: 0.9935 - recall: 1.0000 - val_loss: 0.5096 - val_acc: 0.9184 - val_auc: 0.9935 - val_recall: 1.0000\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00016: early stopping\n",
      "training time: 61.258931398391724s\n"
     ]
    }
   ],
   "source": [
    "#model_types = [ 'LSTM_2','LSTM', 'dense_2','dense', 'ens', 'dense_fc','conv_1d']\n",
    "model_types = [ 'LSTM', 'dense_2','dense', 'ens', 'dense_fc','conv_1d']\n",
    "#model_types = ['conv_1d', 'LSTM_2']\n",
    "trainig_reulsts = {}\n",
    "\n",
    "for model_name in model_types:\n",
    "    \n",
    "    trained_models, trained_models_stats, trained_models_time_taken, trained_models_best_epoch = generate_trained_models(model_name)\n",
    "    stats_best_epoch = pd.DataFrame(trained_models_best_epoch, columns=['epoch']).mean()\n",
    "    stats_time_takes = pd.DataFrame(trained_models_time_taken, columns=['time']).mean()\n",
    "    stats_best_stats = pd.DataFrame(trained_models_stats).mean()\n",
    "\n",
    "    trainig_reulsts[model_name] = {'trained_models': trained_models,\n",
    "                                  'trained_models_stats': trained_models_stats,\n",
    "                                  'trained_models_time_taken': trained_models_time_taken,\n",
    "                                  'trained_models_best_epoch': trained_models_best_epoch,\n",
    "                                  'stats_best_epoch': stats_best_epoch,\n",
    "                                  'stats_time_takes': stats_time_takes,\n",
    "                                  'stats_best_stats': stats_best_stats}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_____\n",
      "model: LSTM \n",
      "\n",
      "mean best epoc: [40.]\n",
      "mean training time: [275.56412427]\n",
      "STATS:\n",
      "acc           0.865216\n",
      "auc           0.959082\n",
      "loss          0.676688\n",
      "lr            0.000014\n",
      "recall        0.984079\n",
      "val_acc       0.852708\n",
      "val_auc       0.959223\n",
      "val_loss      0.689645\n",
      "val_recall    0.983930\n",
      "dtype: float64\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/javascript": [
       "/* Put everything inside the global mpl namespace */\n",
       "window.mpl = {};\n",
       "\n",
       "\n",
       "mpl.get_websocket_type = function() {\n",
       "    if (typeof(WebSocket) !== 'undefined') {\n",
       "        return WebSocket;\n",
       "    } else if (typeof(MozWebSocket) !== 'undefined') {\n",
       "        return MozWebSocket;\n",
       "    } else {\n",
       "        alert('Your browser does not have WebSocket support.' +\n",
       "              'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
       "              'Firefox 4 and 5 are also supported but you ' +\n",
       "              'have to enable WebSockets in about:config.');\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
       "    this.id = figure_id;\n",
       "\n",
       "    this.ws = websocket;\n",
       "\n",
       "    this.supports_binary = (this.ws.binaryType != undefined);\n",
       "\n",
       "    if (!this.supports_binary) {\n",
       "        var warnings = document.getElementById(\"mpl-warnings\");\n",
       "        if (warnings) {\n",
       "            warnings.style.display = 'block';\n",
       "            warnings.textContent = (\n",
       "                \"This browser does not support binary websocket messages. \" +\n",
       "                    \"Performance may be slow.\");\n",
       "        }\n",
       "    }\n",
       "\n",
       "    this.imageObj = new Image();\n",
       "\n",
       "    this.context = undefined;\n",
       "    this.message = undefined;\n",
       "    this.canvas = undefined;\n",
       "    this.rubberband_canvas = undefined;\n",
       "    this.rubberband_context = undefined;\n",
       "    this.format_dropdown = undefined;\n",
       "\n",
       "    this.image_mode = 'full';\n",
       "\n",
       "    this.root = $('<div/>');\n",
       "    this._root_extra_style(this.root)\n",
       "    this.root.attr('style', 'display: inline-block');\n",
       "\n",
       "    $(parent_element).append(this.root);\n",
       "\n",
       "    this._init_header(this);\n",
       "    this._init_canvas(this);\n",
       "    this._init_toolbar(this);\n",
       "\n",
       "    var fig = this;\n",
       "\n",
       "    this.waiting = false;\n",
       "\n",
       "    this.ws.onopen =  function () {\n",
       "            fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
       "            fig.send_message(\"send_image_mode\", {});\n",
       "            if (mpl.ratio != 1) {\n",
       "                fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
       "            }\n",
       "            fig.send_message(\"refresh\", {});\n",
       "        }\n",
       "\n",
       "    this.imageObj.onload = function() {\n",
       "            if (fig.image_mode == 'full') {\n",
       "                // Full images could contain transparency (where diff images\n",
       "                // almost always do), so we need to clear the canvas so that\n",
       "                // there is no ghosting.\n",
       "                fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
       "            }\n",
       "            fig.context.drawImage(fig.imageObj, 0, 0);\n",
       "        };\n",
       "\n",
       "    this.imageObj.onunload = function() {\n",
       "        fig.ws.close();\n",
       "    }\n",
       "\n",
       "    this.ws.onmessage = this._make_on_message_function(this);\n",
       "\n",
       "    this.ondownload = ondownload;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_header = function() {\n",
       "    var titlebar = $(\n",
       "        '<div class=\"ui-dialog-titlebar ui-widget-header ui-corner-all ' +\n",
       "        'ui-helper-clearfix\"/>');\n",
       "    var titletext = $(\n",
       "        '<div class=\"ui-dialog-title\" style=\"width: 100%; ' +\n",
       "        'text-align: center; padding: 3px;\"/>');\n",
       "    titlebar.append(titletext)\n",
       "    this.root.append(titlebar);\n",
       "    this.header = titletext[0];\n",
       "}\n",
       "\n",
       "\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_canvas = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var canvas_div = $('<div/>');\n",
       "\n",
       "    canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
       "\n",
       "    function canvas_keyboard_event(event) {\n",
       "        return fig.key_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    canvas_div.keydown('key_press', canvas_keyboard_event);\n",
       "    canvas_div.keyup('key_release', canvas_keyboard_event);\n",
       "    this.canvas_div = canvas_div\n",
       "    this._canvas_extra_style(canvas_div)\n",
       "    this.root.append(canvas_div);\n",
       "\n",
       "    var canvas = $('<canvas/>');\n",
       "    canvas.addClass('mpl-canvas');\n",
       "    canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
       "\n",
       "    this.canvas = canvas[0];\n",
       "    this.context = canvas[0].getContext(\"2d\");\n",
       "\n",
       "    var backingStore = this.context.backingStorePixelRatio ||\n",
       "\tthis.context.webkitBackingStorePixelRatio ||\n",
       "\tthis.context.mozBackingStorePixelRatio ||\n",
       "\tthis.context.msBackingStorePixelRatio ||\n",
       "\tthis.context.oBackingStorePixelRatio ||\n",
       "\tthis.context.backingStorePixelRatio || 1;\n",
       "\n",
       "    mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
       "\n",
       "    var rubberband = $('<canvas/>');\n",
       "    rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
       "\n",
       "    var pass_mouse_events = true;\n",
       "\n",
       "    canvas_div.resizable({\n",
       "        start: function(event, ui) {\n",
       "            pass_mouse_events = false;\n",
       "        },\n",
       "        resize: function(event, ui) {\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "        stop: function(event, ui) {\n",
       "            pass_mouse_events = true;\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "    });\n",
       "\n",
       "    function mouse_event_fn(event) {\n",
       "        if (pass_mouse_events)\n",
       "            return fig.mouse_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    rubberband.mousedown('button_press', mouse_event_fn);\n",
       "    rubberband.mouseup('button_release', mouse_event_fn);\n",
       "    // Throttle sequential mouse events to 1 every 20ms.\n",
       "    rubberband.mousemove('motion_notify', mouse_event_fn);\n",
       "\n",
       "    rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
       "    rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
       "\n",
       "    canvas_div.on(\"wheel\", function (event) {\n",
       "        event = event.originalEvent;\n",
       "        event['data'] = 'scroll'\n",
       "        if (event.deltaY < 0) {\n",
       "            event.step = 1;\n",
       "        } else {\n",
       "            event.step = -1;\n",
       "        }\n",
       "        mouse_event_fn(event);\n",
       "    });\n",
       "\n",
       "    canvas_div.append(canvas);\n",
       "    canvas_div.append(rubberband);\n",
       "\n",
       "    this.rubberband = rubberband;\n",
       "    this.rubberband_canvas = rubberband[0];\n",
       "    this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
       "    this.rubberband_context.strokeStyle = \"#000000\";\n",
       "\n",
       "    this._resize_canvas = function(width, height) {\n",
       "        // Keep the size of the canvas, canvas container, and rubber band\n",
       "        // canvas in synch.\n",
       "        canvas_div.css('width', width)\n",
       "        canvas_div.css('height', height)\n",
       "\n",
       "        canvas.attr('width', width * mpl.ratio);\n",
       "        canvas.attr('height', height * mpl.ratio);\n",
       "        canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
       "\n",
       "        rubberband.attr('width', width);\n",
       "        rubberband.attr('height', height);\n",
       "    }\n",
       "\n",
       "    // Set the figure to an initial 600x600px, this will subsequently be updated\n",
       "    // upon first draw.\n",
       "    this._resize_canvas(600, 600);\n",
       "\n",
       "    // Disable right mouse context menu.\n",
       "    $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
       "        return false;\n",
       "    });\n",
       "\n",
       "    function set_focus () {\n",
       "        canvas.focus();\n",
       "        canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    window.setTimeout(set_focus, 100);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>')\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items) {\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) {\n",
       "            // put a spacer in here.\n",
       "            continue;\n",
       "        }\n",
       "        var button = $('<button/>');\n",
       "        button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
       "                        'ui-button-icon-only');\n",
       "        button.attr('role', 'button');\n",
       "        button.attr('aria-disabled', 'false');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "\n",
       "        var icon_img = $('<span/>');\n",
       "        icon_img.addClass('ui-button-icon-primary ui-icon');\n",
       "        icon_img.addClass(image);\n",
       "        icon_img.addClass('ui-corner-all');\n",
       "\n",
       "        var tooltip_span = $('<span/>');\n",
       "        tooltip_span.addClass('ui-button-text');\n",
       "        tooltip_span.html(tooltip);\n",
       "\n",
       "        button.append(icon_img);\n",
       "        button.append(tooltip_span);\n",
       "\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    var fmt_picker_span = $('<span/>');\n",
       "\n",
       "    var fmt_picker = $('<select/>');\n",
       "    fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
       "    fmt_picker_span.append(fmt_picker);\n",
       "    nav_element.append(fmt_picker_span);\n",
       "    this.format_dropdown = fmt_picker[0];\n",
       "\n",
       "    for (var ind in mpl.extensions) {\n",
       "        var fmt = mpl.extensions[ind];\n",
       "        var option = $(\n",
       "            '<option/>', {selected: fmt === mpl.default_extension}).html(fmt);\n",
       "        fmt_picker.append(option)\n",
       "    }\n",
       "\n",
       "    // Add hover states to the ui-buttons\n",
       "    $( \".ui-button\" ).hover(\n",
       "        function() { $(this).addClass(\"ui-state-hover\");},\n",
       "        function() { $(this).removeClass(\"ui-state-hover\");}\n",
       "    );\n",
       "\n",
       "    var status_bar = $('<span class=\"mpl-message\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
       "    // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
       "    // which will in turn request a refresh of the image.\n",
       "    this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_message = function(type, properties) {\n",
       "    properties['type'] = type;\n",
       "    properties['figure_id'] = this.id;\n",
       "    this.ws.send(JSON.stringify(properties));\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_draw_message = function() {\n",
       "    if (!this.waiting) {\n",
       "        this.waiting = true;\n",
       "        this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
       "    }\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    var format_dropdown = fig.format_dropdown;\n",
       "    var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
       "    fig.ondownload(fig, format);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
       "    var size = msg['size'];\n",
       "    if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
       "        fig._resize_canvas(size[0], size[1]);\n",
       "        fig.send_message(\"refresh\", {});\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
       "    var x0 = msg['x0'] / mpl.ratio;\n",
       "    var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
       "    var x1 = msg['x1'] / mpl.ratio;\n",
       "    var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
       "    x0 = Math.floor(x0) + 0.5;\n",
       "    y0 = Math.floor(y0) + 0.5;\n",
       "    x1 = Math.floor(x1) + 0.5;\n",
       "    y1 = Math.floor(y1) + 0.5;\n",
       "    var min_x = Math.min(x0, x1);\n",
       "    var min_y = Math.min(y0, y1);\n",
       "    var width = Math.abs(x1 - x0);\n",
       "    var height = Math.abs(y1 - y0);\n",
       "\n",
       "    fig.rubberband_context.clearRect(\n",
       "        0, 0, fig.canvas.width, fig.canvas.height);\n",
       "\n",
       "    fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
       "    // Updates the figure title.\n",
       "    fig.header.textContent = msg['label'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
       "    var cursor = msg['cursor'];\n",
       "    switch(cursor)\n",
       "    {\n",
       "    case 0:\n",
       "        cursor = 'pointer';\n",
       "        break;\n",
       "    case 1:\n",
       "        cursor = 'default';\n",
       "        break;\n",
       "    case 2:\n",
       "        cursor = 'crosshair';\n",
       "        break;\n",
       "    case 3:\n",
       "        cursor = 'move';\n",
       "        break;\n",
       "    }\n",
       "    fig.rubberband_canvas.style.cursor = cursor;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_message = function(fig, msg) {\n",
       "    fig.message.textContent = msg['message'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
       "    // Request the server to send over a new figure.\n",
       "    fig.send_draw_message();\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
       "    fig.image_mode = msg['mode'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Called whenever the canvas gets updated.\n",
       "    this.send_message(\"ack\", {});\n",
       "}\n",
       "\n",
       "// A function to construct a web socket function for onmessage handling.\n",
       "// Called in the figure constructor.\n",
       "mpl.figure.prototype._make_on_message_function = function(fig) {\n",
       "    return function socket_on_message(evt) {\n",
       "        if (evt.data instanceof Blob) {\n",
       "            /* FIXME: We get \"Resource interpreted as Image but\n",
       "             * transferred with MIME type text/plain:\" errors on\n",
       "             * Chrome.  But how to set the MIME type?  It doesn't seem\n",
       "             * to be part of the websocket stream */\n",
       "            evt.data.type = \"image/png\";\n",
       "\n",
       "            /* Free the memory for the previous frames */\n",
       "            if (fig.imageObj.src) {\n",
       "                (window.URL || window.webkitURL).revokeObjectURL(\n",
       "                    fig.imageObj.src);\n",
       "            }\n",
       "\n",
       "            fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
       "                evt.data);\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "        else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
       "            fig.imageObj.src = evt.data;\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        var msg = JSON.parse(evt.data);\n",
       "        var msg_type = msg['type'];\n",
       "\n",
       "        // Call the  \"handle_{type}\" callback, which takes\n",
       "        // the figure and JSON message as its only arguments.\n",
       "        try {\n",
       "            var callback = fig[\"handle_\" + msg_type];\n",
       "        } catch (e) {\n",
       "            console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        if (callback) {\n",
       "            try {\n",
       "                // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
       "                callback(fig, msg);\n",
       "            } catch (e) {\n",
       "                console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
       "            }\n",
       "        }\n",
       "    };\n",
       "}\n",
       "\n",
       "// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
       "mpl.findpos = function(e) {\n",
       "    //this section is from http://www.quirksmode.org/js/events_properties.html\n",
       "    var targ;\n",
       "    if (!e)\n",
       "        e = window.event;\n",
       "    if (e.target)\n",
       "        targ = e.target;\n",
       "    else if (e.srcElement)\n",
       "        targ = e.srcElement;\n",
       "    if (targ.nodeType == 3) // defeat Safari bug\n",
       "        targ = targ.parentNode;\n",
       "\n",
       "    // jQuery normalizes the pageX and pageY\n",
       "    // pageX,Y are the mouse positions relative to the document\n",
       "    // offset() returns the position of the element relative to the document\n",
       "    var x = e.pageX - $(targ).offset().left;\n",
       "    var y = e.pageY - $(targ).offset().top;\n",
       "\n",
       "    return {\"x\": x, \"y\": y};\n",
       "};\n",
       "\n",
       "/*\n",
       " * return a copy of an object with only non-object keys\n",
       " * we need this to avoid circular references\n",
       " * http://stackoverflow.com/a/24161582/3208463\n",
       " */\n",
       "function simpleKeys (original) {\n",
       "  return Object.keys(original).reduce(function (obj, key) {\n",
       "    if (typeof original[key] !== 'object')\n",
       "        obj[key] = original[key]\n",
       "    return obj;\n",
       "  }, {});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.mouse_event = function(event, name) {\n",
       "    var canvas_pos = mpl.findpos(event)\n",
       "\n",
       "    if (name === 'button_press')\n",
       "    {\n",
       "        this.canvas.focus();\n",
       "        this.canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    var x = canvas_pos.x * mpl.ratio;\n",
       "    var y = canvas_pos.y * mpl.ratio;\n",
       "\n",
       "    this.send_message(name, {x: x, y: y, button: event.button,\n",
       "                             step: event.step,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "\n",
       "    /* This prevents the web browser from automatically changing to\n",
       "     * the text insertion cursor when the button is pressed.  We want\n",
       "     * to control all of the cursor setting manually through the\n",
       "     * 'cursor' event from matplotlib */\n",
       "    event.preventDefault();\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    // Handle any extra behaviour associated with a key event\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.key_event = function(event, name) {\n",
       "\n",
       "    // Prevent repeat events\n",
       "    if (name == 'key_press')\n",
       "    {\n",
       "        if (event.which === this._key)\n",
       "            return;\n",
       "        else\n",
       "            this._key = event.which;\n",
       "    }\n",
       "    if (name == 'key_release')\n",
       "        this._key = null;\n",
       "\n",
       "    var value = '';\n",
       "    if (event.ctrlKey && event.which != 17)\n",
       "        value += \"ctrl+\";\n",
       "    if (event.altKey && event.which != 18)\n",
       "        value += \"alt+\";\n",
       "    if (event.shiftKey && event.which != 16)\n",
       "        value += \"shift+\";\n",
       "\n",
       "    value += 'k';\n",
       "    value += event.which.toString();\n",
       "\n",
       "    this._key_event_extra(event, name);\n",
       "\n",
       "    this.send_message(name, {key: value,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
       "    if (name == 'download') {\n",
       "        this.handle_save(this, null);\n",
       "    } else {\n",
       "        this.send_message(\"toolbar_button\", {name: name});\n",
       "    }\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
       "    this.message.textContent = tooltip;\n",
       "};\n",
       "mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
       "\n",
       "mpl.extensions = [\"eps\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\"];\n",
       "\n",
       "mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
       "    // Create a \"websocket\"-like object which calls the given IPython comm\n",
       "    // object with the appropriate methods. Currently this is a non binary\n",
       "    // socket, so there is still some room for performance tuning.\n",
       "    var ws = {};\n",
       "\n",
       "    ws.close = function() {\n",
       "        comm.close()\n",
       "    };\n",
       "    ws.send = function(m) {\n",
       "        //console.log('sending', m);\n",
       "        comm.send(m);\n",
       "    };\n",
       "    // Register the callback with on_msg.\n",
       "    comm.on_msg(function(msg) {\n",
       "        //console.log('receiving', msg['content']['data'], msg);\n",
       "        // Pass the mpl event to the overridden (by mpl) onmessage function.\n",
       "        ws.onmessage(msg['content']['data'])\n",
       "    });\n",
       "    return ws;\n",
       "}\n",
       "\n",
       "mpl.mpl_figure_comm = function(comm, msg) {\n",
       "    // This is the function which gets called when the mpl process\n",
       "    // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
       "\n",
       "    var id = msg.content.data.id;\n",
       "    // Get hold of the div created by the display call when the Comm\n",
       "    // socket was opened in Python.\n",
       "    var element = $(\"#\" + id);\n",
       "    var ws_proxy = comm_websocket_adapter(comm)\n",
       "\n",
       "    function ondownload(figure, format) {\n",
       "        window.open(figure.imageObj.src);\n",
       "    }\n",
       "\n",
       "    var fig = new mpl.figure(id, ws_proxy,\n",
       "                           ondownload,\n",
       "                           element.get(0));\n",
       "\n",
       "    // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
       "    // web socket which is closed, not our websocket->open comm proxy.\n",
       "    ws_proxy.onopen();\n",
       "\n",
       "    fig.parent_element = element.get(0);\n",
       "    fig.cell_info = mpl.find_output_cell(\"<div id='\" + id + \"'></div>\");\n",
       "    if (!fig.cell_info) {\n",
       "        console.error(\"Failed to find cell for figure\", id, fig);\n",
       "        return;\n",
       "    }\n",
       "\n",
       "    var output_index = fig.cell_info[2]\n",
       "    var cell = fig.cell_info[0];\n",
       "\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.handle_close = function(fig, msg) {\n",
       "    var width = fig.canvas.width/mpl.ratio\n",
       "    fig.root.unbind('remove')\n",
       "\n",
       "    // Update the output cell to use the data from the current canvas.\n",
       "    fig.push_to_output();\n",
       "    var dataURL = fig.canvas.toDataURL();\n",
       "    // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
       "    // the notebook keyboard shortcuts fail.\n",
       "    IPython.keyboard_manager.enable()\n",
       "    $(fig.parent_element).html('<img src=\"' + dataURL + '\" width=\"' + width + '\">');\n",
       "    fig.close_ws(fig, msg);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.close_ws = function(fig, msg){\n",
       "    fig.send_message('closing', msg);\n",
       "    // fig.ws.close()\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
       "    // Turn the data on the canvas into data in the output cell.\n",
       "    var width = this.canvas.width/mpl.ratio\n",
       "    var dataURL = this.canvas.toDataURL();\n",
       "    this.cell_info[1]['text/html'] = '<img src=\"' + dataURL + '\" width=\"' + width + '\">';\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Tell IPython that the notebook contents must change.\n",
       "    IPython.notebook.set_dirty(true);\n",
       "    this.send_message(\"ack\", {});\n",
       "    var fig = this;\n",
       "    // Wait a second, then push the new image to the DOM so\n",
       "    // that it is saved nicely (might be nice to debounce this).\n",
       "    setTimeout(function () { fig.push_to_output() }, 1000);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>')\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items){\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) { continue; };\n",
       "\n",
       "        var button = $('<button class=\"btn btn-default\" href=\"#\" title=\"' + name + '\"><i class=\"fa ' + image + ' fa-lg\"></i></button>');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    // Add the status bar.\n",
       "    var status_bar = $('<span class=\"mpl-message\" style=\"text-align:right; float: right;\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "\n",
       "    // Add the close button to the window.\n",
       "    var buttongrp = $('<div class=\"btn-group inline pull-right\"></div>');\n",
       "    var button = $('<button class=\"btn btn-mini btn-primary\" href=\"#\" title=\"Stop Interaction\"><i class=\"fa fa-power-off icon-remove icon-large\"></i></button>');\n",
       "    button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
       "    button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
       "    buttongrp.append(button);\n",
       "    var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
       "    titlebar.prepend(buttongrp);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(el){\n",
       "    var fig = this\n",
       "    el.on(\"remove\", function(){\n",
       "\tfig.close_ws(fig, {});\n",
       "    });\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(el){\n",
       "    // this is important to make the div 'focusable\n",
       "    el.attr('tabindex', 0)\n",
       "    // reach out to IPython and tell the keyboard manager to turn it's self\n",
       "    // off when our div gets focus\n",
       "\n",
       "    // location in version 3\n",
       "    if (IPython.notebook.keyboard_manager) {\n",
       "        IPython.notebook.keyboard_manager.register_events(el);\n",
       "    }\n",
       "    else {\n",
       "        // location in version 2\n",
       "        IPython.keyboard_manager.register_events(el);\n",
       "    }\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    var manager = IPython.notebook.keyboard_manager;\n",
       "    if (!manager)\n",
       "        manager = IPython.keyboard_manager;\n",
       "\n",
       "    // Check for shift+enter\n",
       "    if (event.shiftKey && event.which == 13) {\n",
       "        this.canvas_div.blur();\n",
       "        event.shiftKey = false;\n",
       "        // Send a \"J\" for go to next cell\n",
       "        event.which = 74;\n",
       "        event.keyCode = 74;\n",
       "        manager.command_mode();\n",
       "        manager.handle_keydown(event);\n",
       "    }\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    fig.ondownload(fig, null);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.find_output_cell = function(html_output) {\n",
       "    // Return the cell and output element which can be found *uniquely* in the notebook.\n",
       "    // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
       "    // IPython event is triggered only after the cells have been serialised, which for\n",
       "    // our purposes (turning an active figure into a static one), is too late.\n",
       "    var cells = IPython.notebook.get_cells();\n",
       "    var ncells = cells.length;\n",
       "    for (var i=0; i<ncells; i++) {\n",
       "        var cell = cells[i];\n",
       "        if (cell.cell_type === 'code'){\n",
       "            for (var j=0; j<cell.output_area.outputs.length; j++) {\n",
       "                var data = cell.output_area.outputs[j];\n",
       "                if (data.data) {\n",
       "                    // IPython >= 3 moved mimebundle to data attribute of output\n",
       "                    data = data.data;\n",
       "                }\n",
       "                if (data['text/html'] == html_output) {\n",
       "                    return [cell, data, j];\n",
       "                }\n",
       "            }\n",
       "        }\n",
       "    }\n",
       "}\n",
       "\n",
       "// Register the function which deals with the matplotlib target/channel.\n",
       "// The kernel may be null if the page has been refreshed.\n",
       "if (IPython.notebook.kernel != null) {\n",
       "    IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
       "}\n"
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<img src=\"data:image/png;base64,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\" width=\"400\">"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_____\n",
      "model: dense_2 \n",
      "\n",
      "mean best epoc: [20.33333333]\n",
      "mean training time: [19.49054233]\n",
      "STATS:\n",
      "acc           0.966106\n",
      "auc           0.993378\n",
      "loss          0.077036\n",
      "lr            0.000052\n",
      "recall        1.000000\n",
      "val_acc       0.926807\n",
      "val_auc       0.993440\n",
      "val_loss      0.321176\n",
      "val_recall    1.000000\n",
      "dtype: float64\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/javascript": [
       "/* Put everything inside the global mpl namespace */\n",
       "window.mpl = {};\n",
       "\n",
       "\n",
       "mpl.get_websocket_type = function() {\n",
       "    if (typeof(WebSocket) !== 'undefined') {\n",
       "        return WebSocket;\n",
       "    } else if (typeof(MozWebSocket) !== 'undefined') {\n",
       "        return MozWebSocket;\n",
       "    } else {\n",
       "        alert('Your browser does not have WebSocket support.' +\n",
       "              'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
       "              'Firefox 4 and 5 are also supported but you ' +\n",
       "              'have to enable WebSockets in about:config.');\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
       "    this.id = figure_id;\n",
       "\n",
       "    this.ws = websocket;\n",
       "\n",
       "    this.supports_binary = (this.ws.binaryType != undefined);\n",
       "\n",
       "    if (!this.supports_binary) {\n",
       "        var warnings = document.getElementById(\"mpl-warnings\");\n",
       "        if (warnings) {\n",
       "            warnings.style.display = 'block';\n",
       "            warnings.textContent = (\n",
       "                \"This browser does not support binary websocket messages. \" +\n",
       "                    \"Performance may be slow.\");\n",
       "        }\n",
       "    }\n",
       "\n",
       "    this.imageObj = new Image();\n",
       "\n",
       "    this.context = undefined;\n",
       "    this.message = undefined;\n",
       "    this.canvas = undefined;\n",
       "    this.rubberband_canvas = undefined;\n",
       "    this.rubberband_context = undefined;\n",
       "    this.format_dropdown = undefined;\n",
       "\n",
       "    this.image_mode = 'full';\n",
       "\n",
       "    this.root = $('<div/>');\n",
       "    this._root_extra_style(this.root)\n",
       "    this.root.attr('style', 'display: inline-block');\n",
       "\n",
       "    $(parent_element).append(this.root);\n",
       "\n",
       "    this._init_header(this);\n",
       "    this._init_canvas(this);\n",
       "    this._init_toolbar(this);\n",
       "\n",
       "    var fig = this;\n",
       "\n",
       "    this.waiting = false;\n",
       "\n",
       "    this.ws.onopen =  function () {\n",
       "            fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
       "            fig.send_message(\"send_image_mode\", {});\n",
       "            if (mpl.ratio != 1) {\n",
       "                fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
       "            }\n",
       "            fig.send_message(\"refresh\", {});\n",
       "        }\n",
       "\n",
       "    this.imageObj.onload = function() {\n",
       "            if (fig.image_mode == 'full') {\n",
       "                // Full images could contain transparency (where diff images\n",
       "                // almost always do), so we need to clear the canvas so that\n",
       "                // there is no ghosting.\n",
       "                fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
       "            }\n",
       "            fig.context.drawImage(fig.imageObj, 0, 0);\n",
       "        };\n",
       "\n",
       "    this.imageObj.onunload = function() {\n",
       "        fig.ws.close();\n",
       "    }\n",
       "\n",
       "    this.ws.onmessage = this._make_on_message_function(this);\n",
       "\n",
       "    this.ondownload = ondownload;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_header = function() {\n",
       "    var titlebar = $(\n",
       "        '<div class=\"ui-dialog-titlebar ui-widget-header ui-corner-all ' +\n",
       "        'ui-helper-clearfix\"/>');\n",
       "    var titletext = $(\n",
       "        '<div class=\"ui-dialog-title\" style=\"width: 100%; ' +\n",
       "        'text-align: center; padding: 3px;\"/>');\n",
       "    titlebar.append(titletext)\n",
       "    this.root.append(titlebar);\n",
       "    this.header = titletext[0];\n",
       "}\n",
       "\n",
       "\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_canvas = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var canvas_div = $('<div/>');\n",
       "\n",
       "    canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
       "\n",
       "    function canvas_keyboard_event(event) {\n",
       "        return fig.key_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    canvas_div.keydown('key_press', canvas_keyboard_event);\n",
       "    canvas_div.keyup('key_release', canvas_keyboard_event);\n",
       "    this.canvas_div = canvas_div\n",
       "    this._canvas_extra_style(canvas_div)\n",
       "    this.root.append(canvas_div);\n",
       "\n",
       "    var canvas = $('<canvas/>');\n",
       "    canvas.addClass('mpl-canvas');\n",
       "    canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
       "\n",
       "    this.canvas = canvas[0];\n",
       "    this.context = canvas[0].getContext(\"2d\");\n",
       "\n",
       "    var backingStore = this.context.backingStorePixelRatio ||\n",
       "\tthis.context.webkitBackingStorePixelRatio ||\n",
       "\tthis.context.mozBackingStorePixelRatio ||\n",
       "\tthis.context.msBackingStorePixelRatio ||\n",
       "\tthis.context.oBackingStorePixelRatio ||\n",
       "\tthis.context.backingStorePixelRatio || 1;\n",
       "\n",
       "    mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
       "\n",
       "    var rubberband = $('<canvas/>');\n",
       "    rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
       "\n",
       "    var pass_mouse_events = true;\n",
       "\n",
       "    canvas_div.resizable({\n",
       "        start: function(event, ui) {\n",
       "            pass_mouse_events = false;\n",
       "        },\n",
       "        resize: function(event, ui) {\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "        stop: function(event, ui) {\n",
       "            pass_mouse_events = true;\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "    });\n",
       "\n",
       "    function mouse_event_fn(event) {\n",
       "        if (pass_mouse_events)\n",
       "            return fig.mouse_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    rubberband.mousedown('button_press', mouse_event_fn);\n",
       "    rubberband.mouseup('button_release', mouse_event_fn);\n",
       "    // Throttle sequential mouse events to 1 every 20ms.\n",
       "    rubberband.mousemove('motion_notify', mouse_event_fn);\n",
       "\n",
       "    rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
       "    rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
       "\n",
       "    canvas_div.on(\"wheel\", function (event) {\n",
       "        event = event.originalEvent;\n",
       "        event['data'] = 'scroll'\n",
       "        if (event.deltaY < 0) {\n",
       "            event.step = 1;\n",
       "        } else {\n",
       "            event.step = -1;\n",
       "        }\n",
       "        mouse_event_fn(event);\n",
       "    });\n",
       "\n",
       "    canvas_div.append(canvas);\n",
       "    canvas_div.append(rubberband);\n",
       "\n",
       "    this.rubberband = rubberband;\n",
       "    this.rubberband_canvas = rubberband[0];\n",
       "    this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
       "    this.rubberband_context.strokeStyle = \"#000000\";\n",
       "\n",
       "    this._resize_canvas = function(width, height) {\n",
       "        // Keep the size of the canvas, canvas container, and rubber band\n",
       "        // canvas in synch.\n",
       "        canvas_div.css('width', width)\n",
       "        canvas_div.css('height', height)\n",
       "\n",
       "        canvas.attr('width', width * mpl.ratio);\n",
       "        canvas.attr('height', height * mpl.ratio);\n",
       "        canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
       "\n",
       "        rubberband.attr('width', width);\n",
       "        rubberband.attr('height', height);\n",
       "    }\n",
       "\n",
       "    // Set the figure to an initial 600x600px, this will subsequently be updated\n",
       "    // upon first draw.\n",
       "    this._resize_canvas(600, 600);\n",
       "\n",
       "    // Disable right mouse context menu.\n",
       "    $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
       "        return false;\n",
       "    });\n",
       "\n",
       "    function set_focus () {\n",
       "        canvas.focus();\n",
       "        canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    window.setTimeout(set_focus, 100);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>')\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items) {\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) {\n",
       "            // put a spacer in here.\n",
       "            continue;\n",
       "        }\n",
       "        var button = $('<button/>');\n",
       "        button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
       "                        'ui-button-icon-only');\n",
       "        button.attr('role', 'button');\n",
       "        button.attr('aria-disabled', 'false');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "\n",
       "        var icon_img = $('<span/>');\n",
       "        icon_img.addClass('ui-button-icon-primary ui-icon');\n",
       "        icon_img.addClass(image);\n",
       "        icon_img.addClass('ui-corner-all');\n",
       "\n",
       "        var tooltip_span = $('<span/>');\n",
       "        tooltip_span.addClass('ui-button-text');\n",
       "        tooltip_span.html(tooltip);\n",
       "\n",
       "        button.append(icon_img);\n",
       "        button.append(tooltip_span);\n",
       "\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    var fmt_picker_span = $('<span/>');\n",
       "\n",
       "    var fmt_picker = $('<select/>');\n",
       "    fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
       "    fmt_picker_span.append(fmt_picker);\n",
       "    nav_element.append(fmt_picker_span);\n",
       "    this.format_dropdown = fmt_picker[0];\n",
       "\n",
       "    for (var ind in mpl.extensions) {\n",
       "        var fmt = mpl.extensions[ind];\n",
       "        var option = $(\n",
       "            '<option/>', {selected: fmt === mpl.default_extension}).html(fmt);\n",
       "        fmt_picker.append(option)\n",
       "    }\n",
       "\n",
       "    // Add hover states to the ui-buttons\n",
       "    $( \".ui-button\" ).hover(\n",
       "        function() { $(this).addClass(\"ui-state-hover\");},\n",
       "        function() { $(this).removeClass(\"ui-state-hover\");}\n",
       "    );\n",
       "\n",
       "    var status_bar = $('<span class=\"mpl-message\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
       "    // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
       "    // which will in turn request a refresh of the image.\n",
       "    this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_message = function(type, properties) {\n",
       "    properties['type'] = type;\n",
       "    properties['figure_id'] = this.id;\n",
       "    this.ws.send(JSON.stringify(properties));\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_draw_message = function() {\n",
       "    if (!this.waiting) {\n",
       "        this.waiting = true;\n",
       "        this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
       "    }\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    var format_dropdown = fig.format_dropdown;\n",
       "    var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
       "    fig.ondownload(fig, format);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
       "    var size = msg['size'];\n",
       "    if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
       "        fig._resize_canvas(size[0], size[1]);\n",
       "        fig.send_message(\"refresh\", {});\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
       "    var x0 = msg['x0'] / mpl.ratio;\n",
       "    var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
       "    var x1 = msg['x1'] / mpl.ratio;\n",
       "    var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
       "    x0 = Math.floor(x0) + 0.5;\n",
       "    y0 = Math.floor(y0) + 0.5;\n",
       "    x1 = Math.floor(x1) + 0.5;\n",
       "    y1 = Math.floor(y1) + 0.5;\n",
       "    var min_x = Math.min(x0, x1);\n",
       "    var min_y = Math.min(y0, y1);\n",
       "    var width = Math.abs(x1 - x0);\n",
       "    var height = Math.abs(y1 - y0);\n",
       "\n",
       "    fig.rubberband_context.clearRect(\n",
       "        0, 0, fig.canvas.width, fig.canvas.height);\n",
       "\n",
       "    fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
       "    // Updates the figure title.\n",
       "    fig.header.textContent = msg['label'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
       "    var cursor = msg['cursor'];\n",
       "    switch(cursor)\n",
       "    {\n",
       "    case 0:\n",
       "        cursor = 'pointer';\n",
       "        break;\n",
       "    case 1:\n",
       "        cursor = 'default';\n",
       "        break;\n",
       "    case 2:\n",
       "        cursor = 'crosshair';\n",
       "        break;\n",
       "    case 3:\n",
       "        cursor = 'move';\n",
       "        break;\n",
       "    }\n",
       "    fig.rubberband_canvas.style.cursor = cursor;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_message = function(fig, msg) {\n",
       "    fig.message.textContent = msg['message'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
       "    // Request the server to send over a new figure.\n",
       "    fig.send_draw_message();\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
       "    fig.image_mode = msg['mode'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Called whenever the canvas gets updated.\n",
       "    this.send_message(\"ack\", {});\n",
       "}\n",
       "\n",
       "// A function to construct a web socket function for onmessage handling.\n",
       "// Called in the figure constructor.\n",
       "mpl.figure.prototype._make_on_message_function = function(fig) {\n",
       "    return function socket_on_message(evt) {\n",
       "        if (evt.data instanceof Blob) {\n",
       "            /* FIXME: We get \"Resource interpreted as Image but\n",
       "             * transferred with MIME type text/plain:\" errors on\n",
       "             * Chrome.  But how to set the MIME type?  It doesn't seem\n",
       "             * to be part of the websocket stream */\n",
       "            evt.data.type = \"image/png\";\n",
       "\n",
       "            /* Free the memory for the previous frames */\n",
       "            if (fig.imageObj.src) {\n",
       "                (window.URL || window.webkitURL).revokeObjectURL(\n",
       "                    fig.imageObj.src);\n",
       "            }\n",
       "\n",
       "            fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
       "                evt.data);\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "        else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
       "            fig.imageObj.src = evt.data;\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        var msg = JSON.parse(evt.data);\n",
       "        var msg_type = msg['type'];\n",
       "\n",
       "        // Call the  \"handle_{type}\" callback, which takes\n",
       "        // the figure and JSON message as its only arguments.\n",
       "        try {\n",
       "            var callback = fig[\"handle_\" + msg_type];\n",
       "        } catch (e) {\n",
       "            console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        if (callback) {\n",
       "            try {\n",
       "                // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
       "                callback(fig, msg);\n",
       "            } catch (e) {\n",
       "                console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
       "            }\n",
       "        }\n",
       "    };\n",
       "}\n",
       "\n",
       "// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
       "mpl.findpos = function(e) {\n",
       "    //this section is from http://www.quirksmode.org/js/events_properties.html\n",
       "    var targ;\n",
       "    if (!e)\n",
       "        e = window.event;\n",
       "    if (e.target)\n",
       "        targ = e.target;\n",
       "    else if (e.srcElement)\n",
       "        targ = e.srcElement;\n",
       "    if (targ.nodeType == 3) // defeat Safari bug\n",
       "        targ = targ.parentNode;\n",
       "\n",
       "    // jQuery normalizes the pageX and pageY\n",
       "    // pageX,Y are the mouse positions relative to the document\n",
       "    // offset() returns the position of the element relative to the document\n",
       "    var x = e.pageX - $(targ).offset().left;\n",
       "    var y = e.pageY - $(targ).offset().top;\n",
       "\n",
       "    return {\"x\": x, \"y\": y};\n",
       "};\n",
       "\n",
       "/*\n",
       " * return a copy of an object with only non-object keys\n",
       " * we need this to avoid circular references\n",
       " * http://stackoverflow.com/a/24161582/3208463\n",
       " */\n",
       "function simpleKeys (original) {\n",
       "  return Object.keys(original).reduce(function (obj, key) {\n",
       "    if (typeof original[key] !== 'object')\n",
       "        obj[key] = original[key]\n",
       "    return obj;\n",
       "  }, {});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.mouse_event = function(event, name) {\n",
       "    var canvas_pos = mpl.findpos(event)\n",
       "\n",
       "    if (name === 'button_press')\n",
       "    {\n",
       "        this.canvas.focus();\n",
       "        this.canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    var x = canvas_pos.x * mpl.ratio;\n",
       "    var y = canvas_pos.y * mpl.ratio;\n",
       "\n",
       "    this.send_message(name, {x: x, y: y, button: event.button,\n",
       "                             step: event.step,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "\n",
       "    /* This prevents the web browser from automatically changing to\n",
       "     * the text insertion cursor when the button is pressed.  We want\n",
       "     * to control all of the cursor setting manually through the\n",
       "     * 'cursor' event from matplotlib */\n",
       "    event.preventDefault();\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    // Handle any extra behaviour associated with a key event\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.key_event = function(event, name) {\n",
       "\n",
       "    // Prevent repeat events\n",
       "    if (name == 'key_press')\n",
       "    {\n",
       "        if (event.which === this._key)\n",
       "            return;\n",
       "        else\n",
       "            this._key = event.which;\n",
       "    }\n",
       "    if (name == 'key_release')\n",
       "        this._key = null;\n",
       "\n",
       "    var value = '';\n",
       "    if (event.ctrlKey && event.which != 17)\n",
       "        value += \"ctrl+\";\n",
       "    if (event.altKey && event.which != 18)\n",
       "        value += \"alt+\";\n",
       "    if (event.shiftKey && event.which != 16)\n",
       "        value += \"shift+\";\n",
       "\n",
       "    value += 'k';\n",
       "    value += event.which.toString();\n",
       "\n",
       "    this._key_event_extra(event, name);\n",
       "\n",
       "    this.send_message(name, {key: value,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
       "    if (name == 'download') {\n",
       "        this.handle_save(this, null);\n",
       "    } else {\n",
       "        this.send_message(\"toolbar_button\", {name: name});\n",
       "    }\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
       "    this.message.textContent = tooltip;\n",
       "};\n",
       "mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
       "\n",
       "mpl.extensions = [\"eps\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\"];\n",
       "\n",
       "mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
       "    // Create a \"websocket\"-like object which calls the given IPython comm\n",
       "    // object with the appropriate methods. Currently this is a non binary\n",
       "    // socket, so there is still some room for performance tuning.\n",
       "    var ws = {};\n",
       "\n",
       "    ws.close = function() {\n",
       "        comm.close()\n",
       "    };\n",
       "    ws.send = function(m) {\n",
       "        //console.log('sending', m);\n",
       "        comm.send(m);\n",
       "    };\n",
       "    // Register the callback with on_msg.\n",
       "    comm.on_msg(function(msg) {\n",
       "        //console.log('receiving', msg['content']['data'], msg);\n",
       "        // Pass the mpl event to the overridden (by mpl) onmessage function.\n",
       "        ws.onmessage(msg['content']['data'])\n",
       "    });\n",
       "    return ws;\n",
       "}\n",
       "\n",
       "mpl.mpl_figure_comm = function(comm, msg) {\n",
       "    // This is the function which gets called when the mpl process\n",
       "    // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
       "\n",
       "    var id = msg.content.data.id;\n",
       "    // Get hold of the div created by the display call when the Comm\n",
       "    // socket was opened in Python.\n",
       "    var element = $(\"#\" + id);\n",
       "    var ws_proxy = comm_websocket_adapter(comm)\n",
       "\n",
       "    function ondownload(figure, format) {\n",
       "        window.open(figure.imageObj.src);\n",
       "    }\n",
       "\n",
       "    var fig = new mpl.figure(id, ws_proxy,\n",
       "                           ondownload,\n",
       "                           element.get(0));\n",
       "\n",
       "    // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
       "    // web socket which is closed, not our websocket->open comm proxy.\n",
       "    ws_proxy.onopen();\n",
       "\n",
       "    fig.parent_element = element.get(0);\n",
       "    fig.cell_info = mpl.find_output_cell(\"<div id='\" + id + \"'></div>\");\n",
       "    if (!fig.cell_info) {\n",
       "        console.error(\"Failed to find cell for figure\", id, fig);\n",
       "        return;\n",
       "    }\n",
       "\n",
       "    var output_index = fig.cell_info[2]\n",
       "    var cell = fig.cell_info[0];\n",
       "\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.handle_close = function(fig, msg) {\n",
       "    var width = fig.canvas.width/mpl.ratio\n",
       "    fig.root.unbind('remove')\n",
       "\n",
       "    // Update the output cell to use the data from the current canvas.\n",
       "    fig.push_to_output();\n",
       "    var dataURL = fig.canvas.toDataURL();\n",
       "    // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
       "    // the notebook keyboard shortcuts fail.\n",
       "    IPython.keyboard_manager.enable()\n",
       "    $(fig.parent_element).html('<img src=\"' + dataURL + '\" width=\"' + width + '\">');\n",
       "    fig.close_ws(fig, msg);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.close_ws = function(fig, msg){\n",
       "    fig.send_message('closing', msg);\n",
       "    // fig.ws.close()\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
       "    // Turn the data on the canvas into data in the output cell.\n",
       "    var width = this.canvas.width/mpl.ratio\n",
       "    var dataURL = this.canvas.toDataURL();\n",
       "    this.cell_info[1]['text/html'] = '<img src=\"' + dataURL + '\" width=\"' + width + '\">';\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Tell IPython that the notebook contents must change.\n",
       "    IPython.notebook.set_dirty(true);\n",
       "    this.send_message(\"ack\", {});\n",
       "    var fig = this;\n",
       "    // Wait a second, then push the new image to the DOM so\n",
       "    // that it is saved nicely (might be nice to debounce this).\n",
       "    setTimeout(function () { fig.push_to_output() }, 1000);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>')\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items){\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) { continue; };\n",
       "\n",
       "        var button = $('<button class=\"btn btn-default\" href=\"#\" title=\"' + name + '\"><i class=\"fa ' + image + ' fa-lg\"></i></button>');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    // Add the status bar.\n",
       "    var status_bar = $('<span class=\"mpl-message\" style=\"text-align:right; float: right;\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "\n",
       "    // Add the close button to the window.\n",
       "    var buttongrp = $('<div class=\"btn-group inline pull-right\"></div>');\n",
       "    var button = $('<button class=\"btn btn-mini btn-primary\" href=\"#\" title=\"Stop Interaction\"><i class=\"fa fa-power-off icon-remove icon-large\"></i></button>');\n",
       "    button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
       "    button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
       "    buttongrp.append(button);\n",
       "    var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
       "    titlebar.prepend(buttongrp);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(el){\n",
       "    var fig = this\n",
       "    el.on(\"remove\", function(){\n",
       "\tfig.close_ws(fig, {});\n",
       "    });\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(el){\n",
       "    // this is important to make the div 'focusable\n",
       "    el.attr('tabindex', 0)\n",
       "    // reach out to IPython and tell the keyboard manager to turn it's self\n",
       "    // off when our div gets focus\n",
       "\n",
       "    // location in version 3\n",
       "    if (IPython.notebook.keyboard_manager) {\n",
       "        IPython.notebook.keyboard_manager.register_events(el);\n",
       "    }\n",
       "    else {\n",
       "        // location in version 2\n",
       "        IPython.keyboard_manager.register_events(el);\n",
       "    }\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    var manager = IPython.notebook.keyboard_manager;\n",
       "    if (!manager)\n",
       "        manager = IPython.keyboard_manager;\n",
       "\n",
       "    // Check for shift+enter\n",
       "    if (event.shiftKey && event.which == 13) {\n",
       "        this.canvas_div.blur();\n",
       "        event.shiftKey = false;\n",
       "        // Send a \"J\" for go to next cell\n",
       "        event.which = 74;\n",
       "        event.keyCode = 74;\n",
       "        manager.command_mode();\n",
       "        manager.handle_keydown(event);\n",
       "    }\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    fig.ondownload(fig, null);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.find_output_cell = function(html_output) {\n",
       "    // Return the cell and output element which can be found *uniquely* in the notebook.\n",
       "    // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
       "    // IPython event is triggered only after the cells have been serialised, which for\n",
       "    // our purposes (turning an active figure into a static one), is too late.\n",
       "    var cells = IPython.notebook.get_cells();\n",
       "    var ncells = cells.length;\n",
       "    for (var i=0; i<ncells; i++) {\n",
       "        var cell = cells[i];\n",
       "        if (cell.cell_type === 'code'){\n",
       "            for (var j=0; j<cell.output_area.outputs.length; j++) {\n",
       "                var data = cell.output_area.outputs[j];\n",
       "                if (data.data) {\n",
       "                    // IPython >= 3 moved mimebundle to data attribute of output\n",
       "                    data = data.data;\n",
       "                }\n",
       "                if (data['text/html'] == html_output) {\n",
       "                    return [cell, data, j];\n",
       "                }\n",
       "            }\n",
       "        }\n",
       "    }\n",
       "}\n",
       "\n",
       "// Register the function which deals with the matplotlib target/channel.\n",
       "// The kernel may be null if the page has been refreshed.\n",
       "if (IPython.notebook.kernel != null) {\n",
       "    IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
       "}\n"
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<img src=\"data:image/png;base64,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\" width=\"400\">"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_____\n",
      "model: dense \n",
      "\n",
      "mean best epoc: [20.66666667]\n",
      "mean training time: [21.13987708]\n",
      "STATS:\n",
      "acc           0.957207\n",
      "auc           0.991995\n",
      "loss          0.106021\n",
      "lr            0.000052\n",
      "recall        1.000000\n",
      "val_acc       0.915250\n",
      "val_auc       0.992079\n",
      "val_loss      0.321964\n",
      "val_recall    1.000000\n",
      "dtype: float64\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/javascript": [
       "/* Put everything inside the global mpl namespace */\n",
       "window.mpl = {};\n",
       "\n",
       "\n",
       "mpl.get_websocket_type = function() {\n",
       "    if (typeof(WebSocket) !== 'undefined') {\n",
       "        return WebSocket;\n",
       "    } else if (typeof(MozWebSocket) !== 'undefined') {\n",
       "        return MozWebSocket;\n",
       "    } else {\n",
       "        alert('Your browser does not have WebSocket support.' +\n",
       "              'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
       "              'Firefox 4 and 5 are also supported but you ' +\n",
       "              'have to enable WebSockets in about:config.');\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
       "    this.id = figure_id;\n",
       "\n",
       "    this.ws = websocket;\n",
       "\n",
       "    this.supports_binary = (this.ws.binaryType != undefined);\n",
       "\n",
       "    if (!this.supports_binary) {\n",
       "        var warnings = document.getElementById(\"mpl-warnings\");\n",
       "        if (warnings) {\n",
       "            warnings.style.display = 'block';\n",
       "            warnings.textContent = (\n",
       "                \"This browser does not support binary websocket messages. \" +\n",
       "                    \"Performance may be slow.\");\n",
       "        }\n",
       "    }\n",
       "\n",
       "    this.imageObj = new Image();\n",
       "\n",
       "    this.context = undefined;\n",
       "    this.message = undefined;\n",
       "    this.canvas = undefined;\n",
       "    this.rubberband_canvas = undefined;\n",
       "    this.rubberband_context = undefined;\n",
       "    this.format_dropdown = undefined;\n",
       "\n",
       "    this.image_mode = 'full';\n",
       "\n",
       "    this.root = $('<div/>');\n",
       "    this._root_extra_style(this.root)\n",
       "    this.root.attr('style', 'display: inline-block');\n",
       "\n",
       "    $(parent_element).append(this.root);\n",
       "\n",
       "    this._init_header(this);\n",
       "    this._init_canvas(this);\n",
       "    this._init_toolbar(this);\n",
       "\n",
       "    var fig = this;\n",
       "\n",
       "    this.waiting = false;\n",
       "\n",
       "    this.ws.onopen =  function () {\n",
       "            fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
       "            fig.send_message(\"send_image_mode\", {});\n",
       "            if (mpl.ratio != 1) {\n",
       "                fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
       "            }\n",
       "            fig.send_message(\"refresh\", {});\n",
       "        }\n",
       "\n",
       "    this.imageObj.onload = function() {\n",
       "            if (fig.image_mode == 'full') {\n",
       "                // Full images could contain transparency (where diff images\n",
       "                // almost always do), so we need to clear the canvas so that\n",
       "                // there is no ghosting.\n",
       "                fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
       "            }\n",
       "            fig.context.drawImage(fig.imageObj, 0, 0);\n",
       "        };\n",
       "\n",
       "    this.imageObj.onunload = function() {\n",
       "        fig.ws.close();\n",
       "    }\n",
       "\n",
       "    this.ws.onmessage = this._make_on_message_function(this);\n",
       "\n",
       "    this.ondownload = ondownload;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_header = function() {\n",
       "    var titlebar = $(\n",
       "        '<div class=\"ui-dialog-titlebar ui-widget-header ui-corner-all ' +\n",
       "        'ui-helper-clearfix\"/>');\n",
       "    var titletext = $(\n",
       "        '<div class=\"ui-dialog-title\" style=\"width: 100%; ' +\n",
       "        'text-align: center; padding: 3px;\"/>');\n",
       "    titlebar.append(titletext)\n",
       "    this.root.append(titlebar);\n",
       "    this.header = titletext[0];\n",
       "}\n",
       "\n",
       "\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_canvas = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var canvas_div = $('<div/>');\n",
       "\n",
       "    canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
       "\n",
       "    function canvas_keyboard_event(event) {\n",
       "        return fig.key_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    canvas_div.keydown('key_press', canvas_keyboard_event);\n",
       "    canvas_div.keyup('key_release', canvas_keyboard_event);\n",
       "    this.canvas_div = canvas_div\n",
       "    this._canvas_extra_style(canvas_div)\n",
       "    this.root.append(canvas_div);\n",
       "\n",
       "    var canvas = $('<canvas/>');\n",
       "    canvas.addClass('mpl-canvas');\n",
       "    canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
       "\n",
       "    this.canvas = canvas[0];\n",
       "    this.context = canvas[0].getContext(\"2d\");\n",
       "\n",
       "    var backingStore = this.context.backingStorePixelRatio ||\n",
       "\tthis.context.webkitBackingStorePixelRatio ||\n",
       "\tthis.context.mozBackingStorePixelRatio ||\n",
       "\tthis.context.msBackingStorePixelRatio ||\n",
       "\tthis.context.oBackingStorePixelRatio ||\n",
       "\tthis.context.backingStorePixelRatio || 1;\n",
       "\n",
       "    mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
       "\n",
       "    var rubberband = $('<canvas/>');\n",
       "    rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
       "\n",
       "    var pass_mouse_events = true;\n",
       "\n",
       "    canvas_div.resizable({\n",
       "        start: function(event, ui) {\n",
       "            pass_mouse_events = false;\n",
       "        },\n",
       "        resize: function(event, ui) {\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "        stop: function(event, ui) {\n",
       "            pass_mouse_events = true;\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "    });\n",
       "\n",
       "    function mouse_event_fn(event) {\n",
       "        if (pass_mouse_events)\n",
       "            return fig.mouse_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    rubberband.mousedown('button_press', mouse_event_fn);\n",
       "    rubberband.mouseup('button_release', mouse_event_fn);\n",
       "    // Throttle sequential mouse events to 1 every 20ms.\n",
       "    rubberband.mousemove('motion_notify', mouse_event_fn);\n",
       "\n",
       "    rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
       "    rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
       "\n",
       "    canvas_div.on(\"wheel\", function (event) {\n",
       "        event = event.originalEvent;\n",
       "        event['data'] = 'scroll'\n",
       "        if (event.deltaY < 0) {\n",
       "            event.step = 1;\n",
       "        } else {\n",
       "            event.step = -1;\n",
       "        }\n",
       "        mouse_event_fn(event);\n",
       "    });\n",
       "\n",
       "    canvas_div.append(canvas);\n",
       "    canvas_div.append(rubberband);\n",
       "\n",
       "    this.rubberband = rubberband;\n",
       "    this.rubberband_canvas = rubberband[0];\n",
       "    this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
       "    this.rubberband_context.strokeStyle = \"#000000\";\n",
       "\n",
       "    this._resize_canvas = function(width, height) {\n",
       "        // Keep the size of the canvas, canvas container, and rubber band\n",
       "        // canvas in synch.\n",
       "        canvas_div.css('width', width)\n",
       "        canvas_div.css('height', height)\n",
       "\n",
       "        canvas.attr('width', width * mpl.ratio);\n",
       "        canvas.attr('height', height * mpl.ratio);\n",
       "        canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
       "\n",
       "        rubberband.attr('width', width);\n",
       "        rubberband.attr('height', height);\n",
       "    }\n",
       "\n",
       "    // Set the figure to an initial 600x600px, this will subsequently be updated\n",
       "    // upon first draw.\n",
       "    this._resize_canvas(600, 600);\n",
       "\n",
       "    // Disable right mouse context menu.\n",
       "    $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
       "        return false;\n",
       "    });\n",
       "\n",
       "    function set_focus () {\n",
       "        canvas.focus();\n",
       "        canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    window.setTimeout(set_focus, 100);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>')\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items) {\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) {\n",
       "            // put a spacer in here.\n",
       "            continue;\n",
       "        }\n",
       "        var button = $('<button/>');\n",
       "        button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
       "                        'ui-button-icon-only');\n",
       "        button.attr('role', 'button');\n",
       "        button.attr('aria-disabled', 'false');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "\n",
       "        var icon_img = $('<span/>');\n",
       "        icon_img.addClass('ui-button-icon-primary ui-icon');\n",
       "        icon_img.addClass(image);\n",
       "        icon_img.addClass('ui-corner-all');\n",
       "\n",
       "        var tooltip_span = $('<span/>');\n",
       "        tooltip_span.addClass('ui-button-text');\n",
       "        tooltip_span.html(tooltip);\n",
       "\n",
       "        button.append(icon_img);\n",
       "        button.append(tooltip_span);\n",
       "\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    var fmt_picker_span = $('<span/>');\n",
       "\n",
       "    var fmt_picker = $('<select/>');\n",
       "    fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
       "    fmt_picker_span.append(fmt_picker);\n",
       "    nav_element.append(fmt_picker_span);\n",
       "    this.format_dropdown = fmt_picker[0];\n",
       "\n",
       "    for (var ind in mpl.extensions) {\n",
       "        var fmt = mpl.extensions[ind];\n",
       "        var option = $(\n",
       "            '<option/>', {selected: fmt === mpl.default_extension}).html(fmt);\n",
       "        fmt_picker.append(option)\n",
       "    }\n",
       "\n",
       "    // Add hover states to the ui-buttons\n",
       "    $( \".ui-button\" ).hover(\n",
       "        function() { $(this).addClass(\"ui-state-hover\");},\n",
       "        function() { $(this).removeClass(\"ui-state-hover\");}\n",
       "    );\n",
       "\n",
       "    var status_bar = $('<span class=\"mpl-message\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
       "    // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
       "    // which will in turn request a refresh of the image.\n",
       "    this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_message = function(type, properties) {\n",
       "    properties['type'] = type;\n",
       "    properties['figure_id'] = this.id;\n",
       "    this.ws.send(JSON.stringify(properties));\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_draw_message = function() {\n",
       "    if (!this.waiting) {\n",
       "        this.waiting = true;\n",
       "        this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
       "    }\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    var format_dropdown = fig.format_dropdown;\n",
       "    var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
       "    fig.ondownload(fig, format);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
       "    var size = msg['size'];\n",
       "    if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
       "        fig._resize_canvas(size[0], size[1]);\n",
       "        fig.send_message(\"refresh\", {});\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
       "    var x0 = msg['x0'] / mpl.ratio;\n",
       "    var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
       "    var x1 = msg['x1'] / mpl.ratio;\n",
       "    var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
       "    x0 = Math.floor(x0) + 0.5;\n",
       "    y0 = Math.floor(y0) + 0.5;\n",
       "    x1 = Math.floor(x1) + 0.5;\n",
       "    y1 = Math.floor(y1) + 0.5;\n",
       "    var min_x = Math.min(x0, x1);\n",
       "    var min_y = Math.min(y0, y1);\n",
       "    var width = Math.abs(x1 - x0);\n",
       "    var height = Math.abs(y1 - y0);\n",
       "\n",
       "    fig.rubberband_context.clearRect(\n",
       "        0, 0, fig.canvas.width, fig.canvas.height);\n",
       "\n",
       "    fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
       "    // Updates the figure title.\n",
       "    fig.header.textContent = msg['label'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
       "    var cursor = msg['cursor'];\n",
       "    switch(cursor)\n",
       "    {\n",
       "    case 0:\n",
       "        cursor = 'pointer';\n",
       "        break;\n",
       "    case 1:\n",
       "        cursor = 'default';\n",
       "        break;\n",
       "    case 2:\n",
       "        cursor = 'crosshair';\n",
       "        break;\n",
       "    case 3:\n",
       "        cursor = 'move';\n",
       "        break;\n",
       "    }\n",
       "    fig.rubberband_canvas.style.cursor = cursor;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_message = function(fig, msg) {\n",
       "    fig.message.textContent = msg['message'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
       "    // Request the server to send over a new figure.\n",
       "    fig.send_draw_message();\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
       "    fig.image_mode = msg['mode'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Called whenever the canvas gets updated.\n",
       "    this.send_message(\"ack\", {});\n",
       "}\n",
       "\n",
       "// A function to construct a web socket function for onmessage handling.\n",
       "// Called in the figure constructor.\n",
       "mpl.figure.prototype._make_on_message_function = function(fig) {\n",
       "    return function socket_on_message(evt) {\n",
       "        if (evt.data instanceof Blob) {\n",
       "            /* FIXME: We get \"Resource interpreted as Image but\n",
       "             * transferred with MIME type text/plain:\" errors on\n",
       "             * Chrome.  But how to set the MIME type?  It doesn't seem\n",
       "             * to be part of the websocket stream */\n",
       "            evt.data.type = \"image/png\";\n",
       "\n",
       "            /* Free the memory for the previous frames */\n",
       "            if (fig.imageObj.src) {\n",
       "                (window.URL || window.webkitURL).revokeObjectURL(\n",
       "                    fig.imageObj.src);\n",
       "            }\n",
       "\n",
       "            fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
       "                evt.data);\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "        else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
       "            fig.imageObj.src = evt.data;\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        var msg = JSON.parse(evt.data);\n",
       "        var msg_type = msg['type'];\n",
       "\n",
       "        // Call the  \"handle_{type}\" callback, which takes\n",
       "        // the figure and JSON message as its only arguments.\n",
       "        try {\n",
       "            var callback = fig[\"handle_\" + msg_type];\n",
       "        } catch (e) {\n",
       "            console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        if (callback) {\n",
       "            try {\n",
       "                // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
       "                callback(fig, msg);\n",
       "            } catch (e) {\n",
       "                console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
       "            }\n",
       "        }\n",
       "    };\n",
       "}\n",
       "\n",
       "// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
       "mpl.findpos = function(e) {\n",
       "    //this section is from http://www.quirksmode.org/js/events_properties.html\n",
       "    var targ;\n",
       "    if (!e)\n",
       "        e = window.event;\n",
       "    if (e.target)\n",
       "        targ = e.target;\n",
       "    else if (e.srcElement)\n",
       "        targ = e.srcElement;\n",
       "    if (targ.nodeType == 3) // defeat Safari bug\n",
       "        targ = targ.parentNode;\n",
       "\n",
       "    // jQuery normalizes the pageX and pageY\n",
       "    // pageX,Y are the mouse positions relative to the document\n",
       "    // offset() returns the position of the element relative to the document\n",
       "    var x = e.pageX - $(targ).offset().left;\n",
       "    var y = e.pageY - $(targ).offset().top;\n",
       "\n",
       "    return {\"x\": x, \"y\": y};\n",
       "};\n",
       "\n",
       "/*\n",
       " * return a copy of an object with only non-object keys\n",
       " * we need this to avoid circular references\n",
       " * http://stackoverflow.com/a/24161582/3208463\n",
       " */\n",
       "function simpleKeys (original) {\n",
       "  return Object.keys(original).reduce(function (obj, key) {\n",
       "    if (typeof original[key] !== 'object')\n",
       "        obj[key] = original[key]\n",
       "    return obj;\n",
       "  }, {});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.mouse_event = function(event, name) {\n",
       "    var canvas_pos = mpl.findpos(event)\n",
       "\n",
       "    if (name === 'button_press')\n",
       "    {\n",
       "        this.canvas.focus();\n",
       "        this.canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    var x = canvas_pos.x * mpl.ratio;\n",
       "    var y = canvas_pos.y * mpl.ratio;\n",
       "\n",
       "    this.send_message(name, {x: x, y: y, button: event.button,\n",
       "                             step: event.step,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "\n",
       "    /* This prevents the web browser from automatically changing to\n",
       "     * the text insertion cursor when the button is pressed.  We want\n",
       "     * to control all of the cursor setting manually through the\n",
       "     * 'cursor' event from matplotlib */\n",
       "    event.preventDefault();\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    // Handle any extra behaviour associated with a key event\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.key_event = function(event, name) {\n",
       "\n",
       "    // Prevent repeat events\n",
       "    if (name == 'key_press')\n",
       "    {\n",
       "        if (event.which === this._key)\n",
       "            return;\n",
       "        else\n",
       "            this._key = event.which;\n",
       "    }\n",
       "    if (name == 'key_release')\n",
       "        this._key = null;\n",
       "\n",
       "    var value = '';\n",
       "    if (event.ctrlKey && event.which != 17)\n",
       "        value += \"ctrl+\";\n",
       "    if (event.altKey && event.which != 18)\n",
       "        value += \"alt+\";\n",
       "    if (event.shiftKey && event.which != 16)\n",
       "        value += \"shift+\";\n",
       "\n",
       "    value += 'k';\n",
       "    value += event.which.toString();\n",
       "\n",
       "    this._key_event_extra(event, name);\n",
       "\n",
       "    this.send_message(name, {key: value,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
       "    if (name == 'download') {\n",
       "        this.handle_save(this, null);\n",
       "    } else {\n",
       "        this.send_message(\"toolbar_button\", {name: name});\n",
       "    }\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
       "    this.message.textContent = tooltip;\n",
       "};\n",
       "mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
       "\n",
       "mpl.extensions = [\"eps\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\"];\n",
       "\n",
       "mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
       "    // Create a \"websocket\"-like object which calls the given IPython comm\n",
       "    // object with the appropriate methods. Currently this is a non binary\n",
       "    // socket, so there is still some room for performance tuning.\n",
       "    var ws = {};\n",
       "\n",
       "    ws.close = function() {\n",
       "        comm.close()\n",
       "    };\n",
       "    ws.send = function(m) {\n",
       "        //console.log('sending', m);\n",
       "        comm.send(m);\n",
       "    };\n",
       "    // Register the callback with on_msg.\n",
       "    comm.on_msg(function(msg) {\n",
       "        //console.log('receiving', msg['content']['data'], msg);\n",
       "        // Pass the mpl event to the overridden (by mpl) onmessage function.\n",
       "        ws.onmessage(msg['content']['data'])\n",
       "    });\n",
       "    return ws;\n",
       "}\n",
       "\n",
       "mpl.mpl_figure_comm = function(comm, msg) {\n",
       "    // This is the function which gets called when the mpl process\n",
       "    // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
       "\n",
       "    var id = msg.content.data.id;\n",
       "    // Get hold of the div created by the display call when the Comm\n",
       "    // socket was opened in Python.\n",
       "    var element = $(\"#\" + id);\n",
       "    var ws_proxy = comm_websocket_adapter(comm)\n",
       "\n",
       "    function ondownload(figure, format) {\n",
       "        window.open(figure.imageObj.src);\n",
       "    }\n",
       "\n",
       "    var fig = new mpl.figure(id, ws_proxy,\n",
       "                           ondownload,\n",
       "                           element.get(0));\n",
       "\n",
       "    // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
       "    // web socket which is closed, not our websocket->open comm proxy.\n",
       "    ws_proxy.onopen();\n",
       "\n",
       "    fig.parent_element = element.get(0);\n",
       "    fig.cell_info = mpl.find_output_cell(\"<div id='\" + id + \"'></div>\");\n",
       "    if (!fig.cell_info) {\n",
       "        console.error(\"Failed to find cell for figure\", id, fig);\n",
       "        return;\n",
       "    }\n",
       "\n",
       "    var output_index = fig.cell_info[2]\n",
       "    var cell = fig.cell_info[0];\n",
       "\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.handle_close = function(fig, msg) {\n",
       "    var width = fig.canvas.width/mpl.ratio\n",
       "    fig.root.unbind('remove')\n",
       "\n",
       "    // Update the output cell to use the data from the current canvas.\n",
       "    fig.push_to_output();\n",
       "    var dataURL = fig.canvas.toDataURL();\n",
       "    // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
       "    // the notebook keyboard shortcuts fail.\n",
       "    IPython.keyboard_manager.enable()\n",
       "    $(fig.parent_element).html('<img src=\"' + dataURL + '\" width=\"' + width + '\">');\n",
       "    fig.close_ws(fig, msg);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.close_ws = function(fig, msg){\n",
       "    fig.send_message('closing', msg);\n",
       "    // fig.ws.close()\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
       "    // Turn the data on the canvas into data in the output cell.\n",
       "    var width = this.canvas.width/mpl.ratio\n",
       "    var dataURL = this.canvas.toDataURL();\n",
       "    this.cell_info[1]['text/html'] = '<img src=\"' + dataURL + '\" width=\"' + width + '\">';\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Tell IPython that the notebook contents must change.\n",
       "    IPython.notebook.set_dirty(true);\n",
       "    this.send_message(\"ack\", {});\n",
       "    var fig = this;\n",
       "    // Wait a second, then push the new image to the DOM so\n",
       "    // that it is saved nicely (might be nice to debounce this).\n",
       "    setTimeout(function () { fig.push_to_output() }, 1000);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>')\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items){\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) { continue; };\n",
       "\n",
       "        var button = $('<button class=\"btn btn-default\" href=\"#\" title=\"' + name + '\"><i class=\"fa ' + image + ' fa-lg\"></i></button>');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    // Add the status bar.\n",
       "    var status_bar = $('<span class=\"mpl-message\" style=\"text-align:right; float: right;\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "\n",
       "    // Add the close button to the window.\n",
       "    var buttongrp = $('<div class=\"btn-group inline pull-right\"></div>');\n",
       "    var button = $('<button class=\"btn btn-mini btn-primary\" href=\"#\" title=\"Stop Interaction\"><i class=\"fa fa-power-off icon-remove icon-large\"></i></button>');\n",
       "    button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
       "    button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
       "    buttongrp.append(button);\n",
       "    var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
       "    titlebar.prepend(buttongrp);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(el){\n",
       "    var fig = this\n",
       "    el.on(\"remove\", function(){\n",
       "\tfig.close_ws(fig, {});\n",
       "    });\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(el){\n",
       "    // this is important to make the div 'focusable\n",
       "    el.attr('tabindex', 0)\n",
       "    // reach out to IPython and tell the keyboard manager to turn it's self\n",
       "    // off when our div gets focus\n",
       "\n",
       "    // location in version 3\n",
       "    if (IPython.notebook.keyboard_manager) {\n",
       "        IPython.notebook.keyboard_manager.register_events(el);\n",
       "    }\n",
       "    else {\n",
       "        // location in version 2\n",
       "        IPython.keyboard_manager.register_events(el);\n",
       "    }\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    var manager = IPython.notebook.keyboard_manager;\n",
       "    if (!manager)\n",
       "        manager = IPython.keyboard_manager;\n",
       "\n",
       "    // Check for shift+enter\n",
       "    if (event.shiftKey && event.which == 13) {\n",
       "        this.canvas_div.blur();\n",
       "        event.shiftKey = false;\n",
       "        // Send a \"J\" for go to next cell\n",
       "        event.which = 74;\n",
       "        event.keyCode = 74;\n",
       "        manager.command_mode();\n",
       "        manager.handle_keydown(event);\n",
       "    }\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    fig.ondownload(fig, null);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.find_output_cell = function(html_output) {\n",
       "    // Return the cell and output element which can be found *uniquely* in the notebook.\n",
       "    // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
       "    // IPython event is triggered only after the cells have been serialised, which for\n",
       "    // our purposes (turning an active figure into a static one), is too late.\n",
       "    var cells = IPython.notebook.get_cells();\n",
       "    var ncells = cells.length;\n",
       "    for (var i=0; i<ncells; i++) {\n",
       "        var cell = cells[i];\n",
       "        if (cell.cell_type === 'code'){\n",
       "            for (var j=0; j<cell.output_area.outputs.length; j++) {\n",
       "                var data = cell.output_area.outputs[j];\n",
       "                if (data.data) {\n",
       "                    // IPython >= 3 moved mimebundle to data attribute of output\n",
       "                    data = data.data;\n",
       "                }\n",
       "                if (data['text/html'] == html_output) {\n",
       "                    return [cell, data, j];\n",
       "                }\n",
       "            }\n",
       "        }\n",
       "    }\n",
       "}\n",
       "\n",
       "// Register the function which deals with the matplotlib target/channel.\n",
       "// The kernel may be null if the page has been refreshed.\n",
       "if (IPython.notebook.kernel != null) {\n",
       "    IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
       "}\n"
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<img src=\"data:image/png;base64,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\" width=\"400\">"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_____\n",
      "model: ens \n",
      "\n",
      "mean best epoc: [19.33333333]\n",
      "mean training time: [153.17728448]\n",
      "STATS:\n",
      "acc           0.957093\n",
      "auc           0.989579\n",
      "loss          0.111965\n",
      "lr            0.000063\n",
      "recall        0.996432\n",
      "val_acc       0.905280\n",
      "val_auc       0.989751\n",
      "val_loss      0.310829\n",
      "val_recall    0.996519\n",
      "dtype: float64\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/javascript": [
       "/* Put everything inside the global mpl namespace */\n",
       "window.mpl = {};\n",
       "\n",
       "\n",
       "mpl.get_websocket_type = function() {\n",
       "    if (typeof(WebSocket) !== 'undefined') {\n",
       "        return WebSocket;\n",
       "    } else if (typeof(MozWebSocket) !== 'undefined') {\n",
       "        return MozWebSocket;\n",
       "    } else {\n",
       "        alert('Your browser does not have WebSocket support.' +\n",
       "              'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
       "              'Firefox 4 and 5 are also supported but you ' +\n",
       "              'have to enable WebSockets in about:config.');\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
       "    this.id = figure_id;\n",
       "\n",
       "    this.ws = websocket;\n",
       "\n",
       "    this.supports_binary = (this.ws.binaryType != undefined);\n",
       "\n",
       "    if (!this.supports_binary) {\n",
       "        var warnings = document.getElementById(\"mpl-warnings\");\n",
       "        if (warnings) {\n",
       "            warnings.style.display = 'block';\n",
       "            warnings.textContent = (\n",
       "                \"This browser does not support binary websocket messages. \" +\n",
       "                    \"Performance may be slow.\");\n",
       "        }\n",
       "    }\n",
       "\n",
       "    this.imageObj = new Image();\n",
       "\n",
       "    this.context = undefined;\n",
       "    this.message = undefined;\n",
       "    this.canvas = undefined;\n",
       "    this.rubberband_canvas = undefined;\n",
       "    this.rubberband_context = undefined;\n",
       "    this.format_dropdown = undefined;\n",
       "\n",
       "    this.image_mode = 'full';\n",
       "\n",
       "    this.root = $('<div/>');\n",
       "    this._root_extra_style(this.root)\n",
       "    this.root.attr('style', 'display: inline-block');\n",
       "\n",
       "    $(parent_element).append(this.root);\n",
       "\n",
       "    this._init_header(this);\n",
       "    this._init_canvas(this);\n",
       "    this._init_toolbar(this);\n",
       "\n",
       "    var fig = this;\n",
       "\n",
       "    this.waiting = false;\n",
       "\n",
       "    this.ws.onopen =  function () {\n",
       "            fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
       "            fig.send_message(\"send_image_mode\", {});\n",
       "            if (mpl.ratio != 1) {\n",
       "                fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
       "            }\n",
       "            fig.send_message(\"refresh\", {});\n",
       "        }\n",
       "\n",
       "    this.imageObj.onload = function() {\n",
       "            if (fig.image_mode == 'full') {\n",
       "                // Full images could contain transparency (where diff images\n",
       "                // almost always do), so we need to clear the canvas so that\n",
       "                // there is no ghosting.\n",
       "                fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
       "            }\n",
       "            fig.context.drawImage(fig.imageObj, 0, 0);\n",
       "        };\n",
       "\n",
       "    this.imageObj.onunload = function() {\n",
       "        fig.ws.close();\n",
       "    }\n",
       "\n",
       "    this.ws.onmessage = this._make_on_message_function(this);\n",
       "\n",
       "    this.ondownload = ondownload;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_header = function() {\n",
       "    var titlebar = $(\n",
       "        '<div class=\"ui-dialog-titlebar ui-widget-header ui-corner-all ' +\n",
       "        'ui-helper-clearfix\"/>');\n",
       "    var titletext = $(\n",
       "        '<div class=\"ui-dialog-title\" style=\"width: 100%; ' +\n",
       "        'text-align: center; padding: 3px;\"/>');\n",
       "    titlebar.append(titletext)\n",
       "    this.root.append(titlebar);\n",
       "    this.header = titletext[0];\n",
       "}\n",
       "\n",
       "\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_canvas = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var canvas_div = $('<div/>');\n",
       "\n",
       "    canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
       "\n",
       "    function canvas_keyboard_event(event) {\n",
       "        return fig.key_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    canvas_div.keydown('key_press', canvas_keyboard_event);\n",
       "    canvas_div.keyup('key_release', canvas_keyboard_event);\n",
       "    this.canvas_div = canvas_div\n",
       "    this._canvas_extra_style(canvas_div)\n",
       "    this.root.append(canvas_div);\n",
       "\n",
       "    var canvas = $('<canvas/>');\n",
       "    canvas.addClass('mpl-canvas');\n",
       "    canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
       "\n",
       "    this.canvas = canvas[0];\n",
       "    this.context = canvas[0].getContext(\"2d\");\n",
       "\n",
       "    var backingStore = this.context.backingStorePixelRatio ||\n",
       "\tthis.context.webkitBackingStorePixelRatio ||\n",
       "\tthis.context.mozBackingStorePixelRatio ||\n",
       "\tthis.context.msBackingStorePixelRatio ||\n",
       "\tthis.context.oBackingStorePixelRatio ||\n",
       "\tthis.context.backingStorePixelRatio || 1;\n",
       "\n",
       "    mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
       "\n",
       "    var rubberband = $('<canvas/>');\n",
       "    rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
       "\n",
       "    var pass_mouse_events = true;\n",
       "\n",
       "    canvas_div.resizable({\n",
       "        start: function(event, ui) {\n",
       "            pass_mouse_events = false;\n",
       "        },\n",
       "        resize: function(event, ui) {\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "        stop: function(event, ui) {\n",
       "            pass_mouse_events = true;\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "    });\n",
       "\n",
       "    function mouse_event_fn(event) {\n",
       "        if (pass_mouse_events)\n",
       "            return fig.mouse_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    rubberband.mousedown('button_press', mouse_event_fn);\n",
       "    rubberband.mouseup('button_release', mouse_event_fn);\n",
       "    // Throttle sequential mouse events to 1 every 20ms.\n",
       "    rubberband.mousemove('motion_notify', mouse_event_fn);\n",
       "\n",
       "    rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
       "    rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
       "\n",
       "    canvas_div.on(\"wheel\", function (event) {\n",
       "        event = event.originalEvent;\n",
       "        event['data'] = 'scroll'\n",
       "        if (event.deltaY < 0) {\n",
       "            event.step = 1;\n",
       "        } else {\n",
       "            event.step = -1;\n",
       "        }\n",
       "        mouse_event_fn(event);\n",
       "    });\n",
       "\n",
       "    canvas_div.append(canvas);\n",
       "    canvas_div.append(rubberband);\n",
       "\n",
       "    this.rubberband = rubberband;\n",
       "    this.rubberband_canvas = rubberband[0];\n",
       "    this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
       "    this.rubberband_context.strokeStyle = \"#000000\";\n",
       "\n",
       "    this._resize_canvas = function(width, height) {\n",
       "        // Keep the size of the canvas, canvas container, and rubber band\n",
       "        // canvas in synch.\n",
       "        canvas_div.css('width', width)\n",
       "        canvas_div.css('height', height)\n",
       "\n",
       "        canvas.attr('width', width * mpl.ratio);\n",
       "        canvas.attr('height', height * mpl.ratio);\n",
       "        canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
       "\n",
       "        rubberband.attr('width', width);\n",
       "        rubberband.attr('height', height);\n",
       "    }\n",
       "\n",
       "    // Set the figure to an initial 600x600px, this will subsequently be updated\n",
       "    // upon first draw.\n",
       "    this._resize_canvas(600, 600);\n",
       "\n",
       "    // Disable right mouse context menu.\n",
       "    $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
       "        return false;\n",
       "    });\n",
       "\n",
       "    function set_focus () {\n",
       "        canvas.focus();\n",
       "        canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    window.setTimeout(set_focus, 100);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>')\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items) {\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) {\n",
       "            // put a spacer in here.\n",
       "            continue;\n",
       "        }\n",
       "        var button = $('<button/>');\n",
       "        button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
       "                        'ui-button-icon-only');\n",
       "        button.attr('role', 'button');\n",
       "        button.attr('aria-disabled', 'false');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "\n",
       "        var icon_img = $('<span/>');\n",
       "        icon_img.addClass('ui-button-icon-primary ui-icon');\n",
       "        icon_img.addClass(image);\n",
       "        icon_img.addClass('ui-corner-all');\n",
       "\n",
       "        var tooltip_span = $('<span/>');\n",
       "        tooltip_span.addClass('ui-button-text');\n",
       "        tooltip_span.html(tooltip);\n",
       "\n",
       "        button.append(icon_img);\n",
       "        button.append(tooltip_span);\n",
       "\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    var fmt_picker_span = $('<span/>');\n",
       "\n",
       "    var fmt_picker = $('<select/>');\n",
       "    fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
       "    fmt_picker_span.append(fmt_picker);\n",
       "    nav_element.append(fmt_picker_span);\n",
       "    this.format_dropdown = fmt_picker[0];\n",
       "\n",
       "    for (var ind in mpl.extensions) {\n",
       "        var fmt = mpl.extensions[ind];\n",
       "        var option = $(\n",
       "            '<option/>', {selected: fmt === mpl.default_extension}).html(fmt);\n",
       "        fmt_picker.append(option)\n",
       "    }\n",
       "\n",
       "    // Add hover states to the ui-buttons\n",
       "    $( \".ui-button\" ).hover(\n",
       "        function() { $(this).addClass(\"ui-state-hover\");},\n",
       "        function() { $(this).removeClass(\"ui-state-hover\");}\n",
       "    );\n",
       "\n",
       "    var status_bar = $('<span class=\"mpl-message\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
       "    // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
       "    // which will in turn request a refresh of the image.\n",
       "    this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_message = function(type, properties) {\n",
       "    properties['type'] = type;\n",
       "    properties['figure_id'] = this.id;\n",
       "    this.ws.send(JSON.stringify(properties));\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_draw_message = function() {\n",
       "    if (!this.waiting) {\n",
       "        this.waiting = true;\n",
       "        this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
       "    }\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    var format_dropdown = fig.format_dropdown;\n",
       "    var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
       "    fig.ondownload(fig, format);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
       "    var size = msg['size'];\n",
       "    if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
       "        fig._resize_canvas(size[0], size[1]);\n",
       "        fig.send_message(\"refresh\", {});\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
       "    var x0 = msg['x0'] / mpl.ratio;\n",
       "    var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
       "    var x1 = msg['x1'] / mpl.ratio;\n",
       "    var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
       "    x0 = Math.floor(x0) + 0.5;\n",
       "    y0 = Math.floor(y0) + 0.5;\n",
       "    x1 = Math.floor(x1) + 0.5;\n",
       "    y1 = Math.floor(y1) + 0.5;\n",
       "    var min_x = Math.min(x0, x1);\n",
       "    var min_y = Math.min(y0, y1);\n",
       "    var width = Math.abs(x1 - x0);\n",
       "    var height = Math.abs(y1 - y0);\n",
       "\n",
       "    fig.rubberband_context.clearRect(\n",
       "        0, 0, fig.canvas.width, fig.canvas.height);\n",
       "\n",
       "    fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
       "    // Updates the figure title.\n",
       "    fig.header.textContent = msg['label'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
       "    var cursor = msg['cursor'];\n",
       "    switch(cursor)\n",
       "    {\n",
       "    case 0:\n",
       "        cursor = 'pointer';\n",
       "        break;\n",
       "    case 1:\n",
       "        cursor = 'default';\n",
       "        break;\n",
       "    case 2:\n",
       "        cursor = 'crosshair';\n",
       "        break;\n",
       "    case 3:\n",
       "        cursor = 'move';\n",
       "        break;\n",
       "    }\n",
       "    fig.rubberband_canvas.style.cursor = cursor;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_message = function(fig, msg) {\n",
       "    fig.message.textContent = msg['message'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
       "    // Request the server to send over a new figure.\n",
       "    fig.send_draw_message();\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
       "    fig.image_mode = msg['mode'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Called whenever the canvas gets updated.\n",
       "    this.send_message(\"ack\", {});\n",
       "}\n",
       "\n",
       "// A function to construct a web socket function for onmessage handling.\n",
       "// Called in the figure constructor.\n",
       "mpl.figure.prototype._make_on_message_function = function(fig) {\n",
       "    return function socket_on_message(evt) {\n",
       "        if (evt.data instanceof Blob) {\n",
       "            /* FIXME: We get \"Resource interpreted as Image but\n",
       "             * transferred with MIME type text/plain:\" errors on\n",
       "             * Chrome.  But how to set the MIME type?  It doesn't seem\n",
       "             * to be part of the websocket stream */\n",
       "            evt.data.type = \"image/png\";\n",
       "\n",
       "            /* Free the memory for the previous frames */\n",
       "            if (fig.imageObj.src) {\n",
       "                (window.URL || window.webkitURL).revokeObjectURL(\n",
       "                    fig.imageObj.src);\n",
       "            }\n",
       "\n",
       "            fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
       "                evt.data);\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "        else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
       "            fig.imageObj.src = evt.data;\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        var msg = JSON.parse(evt.data);\n",
       "        var msg_type = msg['type'];\n",
       "\n",
       "        // Call the  \"handle_{type}\" callback, which takes\n",
       "        // the figure and JSON message as its only arguments.\n",
       "        try {\n",
       "            var callback = fig[\"handle_\" + msg_type];\n",
       "        } catch (e) {\n",
       "            console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        if (callback) {\n",
       "            try {\n",
       "                // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
       "                callback(fig, msg);\n",
       "            } catch (e) {\n",
       "                console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
       "            }\n",
       "        }\n",
       "    };\n",
       "}\n",
       "\n",
       "// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
       "mpl.findpos = function(e) {\n",
       "    //this section is from http://www.quirksmode.org/js/events_properties.html\n",
       "    var targ;\n",
       "    if (!e)\n",
       "        e = window.event;\n",
       "    if (e.target)\n",
       "        targ = e.target;\n",
       "    else if (e.srcElement)\n",
       "        targ = e.srcElement;\n",
       "    if (targ.nodeType == 3) // defeat Safari bug\n",
       "        targ = targ.parentNode;\n",
       "\n",
       "    // jQuery normalizes the pageX and pageY\n",
       "    // pageX,Y are the mouse positions relative to the document\n",
       "    // offset() returns the position of the element relative to the document\n",
       "    var x = e.pageX - $(targ).offset().left;\n",
       "    var y = e.pageY - $(targ).offset().top;\n",
       "\n",
       "    return {\"x\": x, \"y\": y};\n",
       "};\n",
       "\n",
       "/*\n",
       " * return a copy of an object with only non-object keys\n",
       " * we need this to avoid circular references\n",
       " * http://stackoverflow.com/a/24161582/3208463\n",
       " */\n",
       "function simpleKeys (original) {\n",
       "  return Object.keys(original).reduce(function (obj, key) {\n",
       "    if (typeof original[key] !== 'object')\n",
       "        obj[key] = original[key]\n",
       "    return obj;\n",
       "  }, {});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.mouse_event = function(event, name) {\n",
       "    var canvas_pos = mpl.findpos(event)\n",
       "\n",
       "    if (name === 'button_press')\n",
       "    {\n",
       "        this.canvas.focus();\n",
       "        this.canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    var x = canvas_pos.x * mpl.ratio;\n",
       "    var y = canvas_pos.y * mpl.ratio;\n",
       "\n",
       "    this.send_message(name, {x: x, y: y, button: event.button,\n",
       "                             step: event.step,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "\n",
       "    /* This prevents the web browser from automatically changing to\n",
       "     * the text insertion cursor when the button is pressed.  We want\n",
       "     * to control all of the cursor setting manually through the\n",
       "     * 'cursor' event from matplotlib */\n",
       "    event.preventDefault();\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    // Handle any extra behaviour associated with a key event\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.key_event = function(event, name) {\n",
       "\n",
       "    // Prevent repeat events\n",
       "    if (name == 'key_press')\n",
       "    {\n",
       "        if (event.which === this._key)\n",
       "            return;\n",
       "        else\n",
       "            this._key = event.which;\n",
       "    }\n",
       "    if (name == 'key_release')\n",
       "        this._key = null;\n",
       "\n",
       "    var value = '';\n",
       "    if (event.ctrlKey && event.which != 17)\n",
       "        value += \"ctrl+\";\n",
       "    if (event.altKey && event.which != 18)\n",
       "        value += \"alt+\";\n",
       "    if (event.shiftKey && event.which != 16)\n",
       "        value += \"shift+\";\n",
       "\n",
       "    value += 'k';\n",
       "    value += event.which.toString();\n",
       "\n",
       "    this._key_event_extra(event, name);\n",
       "\n",
       "    this.send_message(name, {key: value,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
       "    if (name == 'download') {\n",
       "        this.handle_save(this, null);\n",
       "    } else {\n",
       "        this.send_message(\"toolbar_button\", {name: name});\n",
       "    }\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
       "    this.message.textContent = tooltip;\n",
       "};\n",
       "mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
       "\n",
       "mpl.extensions = [\"eps\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\"];\n",
       "\n",
       "mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
       "    // Create a \"websocket\"-like object which calls the given IPython comm\n",
       "    // object with the appropriate methods. Currently this is a non binary\n",
       "    // socket, so there is still some room for performance tuning.\n",
       "    var ws = {};\n",
       "\n",
       "    ws.close = function() {\n",
       "        comm.close()\n",
       "    };\n",
       "    ws.send = function(m) {\n",
       "        //console.log('sending', m);\n",
       "        comm.send(m);\n",
       "    };\n",
       "    // Register the callback with on_msg.\n",
       "    comm.on_msg(function(msg) {\n",
       "        //console.log('receiving', msg['content']['data'], msg);\n",
       "        // Pass the mpl event to the overridden (by mpl) onmessage function.\n",
       "        ws.onmessage(msg['content']['data'])\n",
       "    });\n",
       "    return ws;\n",
       "}\n",
       "\n",
       "mpl.mpl_figure_comm = function(comm, msg) {\n",
       "    // This is the function which gets called when the mpl process\n",
       "    // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
       "\n",
       "    var id = msg.content.data.id;\n",
       "    // Get hold of the div created by the display call when the Comm\n",
       "    // socket was opened in Python.\n",
       "    var element = $(\"#\" + id);\n",
       "    var ws_proxy = comm_websocket_adapter(comm)\n",
       "\n",
       "    function ondownload(figure, format) {\n",
       "        window.open(figure.imageObj.src);\n",
       "    }\n",
       "\n",
       "    var fig = new mpl.figure(id, ws_proxy,\n",
       "                           ondownload,\n",
       "                           element.get(0));\n",
       "\n",
       "    // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
       "    // web socket which is closed, not our websocket->open comm proxy.\n",
       "    ws_proxy.onopen();\n",
       "\n",
       "    fig.parent_element = element.get(0);\n",
       "    fig.cell_info = mpl.find_output_cell(\"<div id='\" + id + \"'></div>\");\n",
       "    if (!fig.cell_info) {\n",
       "        console.error(\"Failed to find cell for figure\", id, fig);\n",
       "        return;\n",
       "    }\n",
       "\n",
       "    var output_index = fig.cell_info[2]\n",
       "    var cell = fig.cell_info[0];\n",
       "\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.handle_close = function(fig, msg) {\n",
       "    var width = fig.canvas.width/mpl.ratio\n",
       "    fig.root.unbind('remove')\n",
       "\n",
       "    // Update the output cell to use the data from the current canvas.\n",
       "    fig.push_to_output();\n",
       "    var dataURL = fig.canvas.toDataURL();\n",
       "    // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
       "    // the notebook keyboard shortcuts fail.\n",
       "    IPython.keyboard_manager.enable()\n",
       "    $(fig.parent_element).html('<img src=\"' + dataURL + '\" width=\"' + width + '\">');\n",
       "    fig.close_ws(fig, msg);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.close_ws = function(fig, msg){\n",
       "    fig.send_message('closing', msg);\n",
       "    // fig.ws.close()\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
       "    // Turn the data on the canvas into data in the output cell.\n",
       "    var width = this.canvas.width/mpl.ratio\n",
       "    var dataURL = this.canvas.toDataURL();\n",
       "    this.cell_info[1]['text/html'] = '<img src=\"' + dataURL + '\" width=\"' + width + '\">';\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Tell IPython that the notebook contents must change.\n",
       "    IPython.notebook.set_dirty(true);\n",
       "    this.send_message(\"ack\", {});\n",
       "    var fig = this;\n",
       "    // Wait a second, then push the new image to the DOM so\n",
       "    // that it is saved nicely (might be nice to debounce this).\n",
       "    setTimeout(function () { fig.push_to_output() }, 1000);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>')\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items){\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) { continue; };\n",
       "\n",
       "        var button = $('<button class=\"btn btn-default\" href=\"#\" title=\"' + name + '\"><i class=\"fa ' + image + ' fa-lg\"></i></button>');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    // Add the status bar.\n",
       "    var status_bar = $('<span class=\"mpl-message\" style=\"text-align:right; float: right;\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "\n",
       "    // Add the close button to the window.\n",
       "    var buttongrp = $('<div class=\"btn-group inline pull-right\"></div>');\n",
       "    var button = $('<button class=\"btn btn-mini btn-primary\" href=\"#\" title=\"Stop Interaction\"><i class=\"fa fa-power-off icon-remove icon-large\"></i></button>');\n",
       "    button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
       "    button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
       "    buttongrp.append(button);\n",
       "    var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
       "    titlebar.prepend(buttongrp);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(el){\n",
       "    var fig = this\n",
       "    el.on(\"remove\", function(){\n",
       "\tfig.close_ws(fig, {});\n",
       "    });\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(el){\n",
       "    // this is important to make the div 'focusable\n",
       "    el.attr('tabindex', 0)\n",
       "    // reach out to IPython and tell the keyboard manager to turn it's self\n",
       "    // off when our div gets focus\n",
       "\n",
       "    // location in version 3\n",
       "    if (IPython.notebook.keyboard_manager) {\n",
       "        IPython.notebook.keyboard_manager.register_events(el);\n",
       "    }\n",
       "    else {\n",
       "        // location in version 2\n",
       "        IPython.keyboard_manager.register_events(el);\n",
       "    }\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    var manager = IPython.notebook.keyboard_manager;\n",
       "    if (!manager)\n",
       "        manager = IPython.keyboard_manager;\n",
       "\n",
       "    // Check for shift+enter\n",
       "    if (event.shiftKey && event.which == 13) {\n",
       "        this.canvas_div.blur();\n",
       "        event.shiftKey = false;\n",
       "        // Send a \"J\" for go to next cell\n",
       "        event.which = 74;\n",
       "        event.keyCode = 74;\n",
       "        manager.command_mode();\n",
       "        manager.handle_keydown(event);\n",
       "    }\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    fig.ondownload(fig, null);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.find_output_cell = function(html_output) {\n",
       "    // Return the cell and output element which can be found *uniquely* in the notebook.\n",
       "    // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
       "    // IPython event is triggered only after the cells have been serialised, which for\n",
       "    // our purposes (turning an active figure into a static one), is too late.\n",
       "    var cells = IPython.notebook.get_cells();\n",
       "    var ncells = cells.length;\n",
       "    for (var i=0; i<ncells; i++) {\n",
       "        var cell = cells[i];\n",
       "        if (cell.cell_type === 'code'){\n",
       "            for (var j=0; j<cell.output_area.outputs.length; j++) {\n",
       "                var data = cell.output_area.outputs[j];\n",
       "                if (data.data) {\n",
       "                    // IPython >= 3 moved mimebundle to data attribute of output\n",
       "                    data = data.data;\n",
       "                }\n",
       "                if (data['text/html'] == html_output) {\n",
       "                    return [cell, data, j];\n",
       "                }\n",
       "            }\n",
       "        }\n",
       "    }\n",
       "}\n",
       "\n",
       "// Register the function which deals with the matplotlib target/channel.\n",
       "// The kernel may be null if the page has been refreshed.\n",
       "if (IPython.notebook.kernel != null) {\n",
       "    IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
       "}\n"
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<img src=\"data:image/png;base64,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\" width=\"400\">"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_____\n",
      "model: dense_fc \n",
      "\n",
      "mean best epoc: [18.66666667]\n",
      "mean training time: [86.19546501]\n",
      "STATS:\n",
      "acc           0.958680\n",
      "auc           0.988381\n",
      "loss          0.101426\n",
      "lr            0.000063\n",
      "recall        1.000000\n",
      "val_acc       0.907999\n",
      "val_auc       0.988512\n",
      "val_loss      0.669495\n",
      "val_recall    1.000000\n",
      "dtype: float64\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/javascript": [
       "/* Put everything inside the global mpl namespace */\n",
       "window.mpl = {};\n",
       "\n",
       "\n",
       "mpl.get_websocket_type = function() {\n",
       "    if (typeof(WebSocket) !== 'undefined') {\n",
       "        return WebSocket;\n",
       "    } else if (typeof(MozWebSocket) !== 'undefined') {\n",
       "        return MozWebSocket;\n",
       "    } else {\n",
       "        alert('Your browser does not have WebSocket support.' +\n",
       "              'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
       "              'Firefox 4 and 5 are also supported but you ' +\n",
       "              'have to enable WebSockets in about:config.');\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
       "    this.id = figure_id;\n",
       "\n",
       "    this.ws = websocket;\n",
       "\n",
       "    this.supports_binary = (this.ws.binaryType != undefined);\n",
       "\n",
       "    if (!this.supports_binary) {\n",
       "        var warnings = document.getElementById(\"mpl-warnings\");\n",
       "        if (warnings) {\n",
       "            warnings.style.display = 'block';\n",
       "            warnings.textContent = (\n",
       "                \"This browser does not support binary websocket messages. \" +\n",
       "                    \"Performance may be slow.\");\n",
       "        }\n",
       "    }\n",
       "\n",
       "    this.imageObj = new Image();\n",
       "\n",
       "    this.context = undefined;\n",
       "    this.message = undefined;\n",
       "    this.canvas = undefined;\n",
       "    this.rubberband_canvas = undefined;\n",
       "    this.rubberband_context = undefined;\n",
       "    this.format_dropdown = undefined;\n",
       "\n",
       "    this.image_mode = 'full';\n",
       "\n",
       "    this.root = $('<div/>');\n",
       "    this._root_extra_style(this.root)\n",
       "    this.root.attr('style', 'display: inline-block');\n",
       "\n",
       "    $(parent_element).append(this.root);\n",
       "\n",
       "    this._init_header(this);\n",
       "    this._init_canvas(this);\n",
       "    this._init_toolbar(this);\n",
       "\n",
       "    var fig = this;\n",
       "\n",
       "    this.waiting = false;\n",
       "\n",
       "    this.ws.onopen =  function () {\n",
       "            fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
       "            fig.send_message(\"send_image_mode\", {});\n",
       "            if (mpl.ratio != 1) {\n",
       "                fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
       "            }\n",
       "            fig.send_message(\"refresh\", {});\n",
       "        }\n",
       "\n",
       "    this.imageObj.onload = function() {\n",
       "            if (fig.image_mode == 'full') {\n",
       "                // Full images could contain transparency (where diff images\n",
       "                // almost always do), so we need to clear the canvas so that\n",
       "                // there is no ghosting.\n",
       "                fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
       "            }\n",
       "            fig.context.drawImage(fig.imageObj, 0, 0);\n",
       "        };\n",
       "\n",
       "    this.imageObj.onunload = function() {\n",
       "        fig.ws.close();\n",
       "    }\n",
       "\n",
       "    this.ws.onmessage = this._make_on_message_function(this);\n",
       "\n",
       "    this.ondownload = ondownload;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_header = function() {\n",
       "    var titlebar = $(\n",
       "        '<div class=\"ui-dialog-titlebar ui-widget-header ui-corner-all ' +\n",
       "        'ui-helper-clearfix\"/>');\n",
       "    var titletext = $(\n",
       "        '<div class=\"ui-dialog-title\" style=\"width: 100%; ' +\n",
       "        'text-align: center; padding: 3px;\"/>');\n",
       "    titlebar.append(titletext)\n",
       "    this.root.append(titlebar);\n",
       "    this.header = titletext[0];\n",
       "}\n",
       "\n",
       "\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_canvas = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var canvas_div = $('<div/>');\n",
       "\n",
       "    canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
       "\n",
       "    function canvas_keyboard_event(event) {\n",
       "        return fig.key_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    canvas_div.keydown('key_press', canvas_keyboard_event);\n",
       "    canvas_div.keyup('key_release', canvas_keyboard_event);\n",
       "    this.canvas_div = canvas_div\n",
       "    this._canvas_extra_style(canvas_div)\n",
       "    this.root.append(canvas_div);\n",
       "\n",
       "    var canvas = $('<canvas/>');\n",
       "    canvas.addClass('mpl-canvas');\n",
       "    canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
       "\n",
       "    this.canvas = canvas[0];\n",
       "    this.context = canvas[0].getContext(\"2d\");\n",
       "\n",
       "    var backingStore = this.context.backingStorePixelRatio ||\n",
       "\tthis.context.webkitBackingStorePixelRatio ||\n",
       "\tthis.context.mozBackingStorePixelRatio ||\n",
       "\tthis.context.msBackingStorePixelRatio ||\n",
       "\tthis.context.oBackingStorePixelRatio ||\n",
       "\tthis.context.backingStorePixelRatio || 1;\n",
       "\n",
       "    mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
       "\n",
       "    var rubberband = $('<canvas/>');\n",
       "    rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
       "\n",
       "    var pass_mouse_events = true;\n",
       "\n",
       "    canvas_div.resizable({\n",
       "        start: function(event, ui) {\n",
       "            pass_mouse_events = false;\n",
       "        },\n",
       "        resize: function(event, ui) {\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "        stop: function(event, ui) {\n",
       "            pass_mouse_events = true;\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "    });\n",
       "\n",
       "    function mouse_event_fn(event) {\n",
       "        if (pass_mouse_events)\n",
       "            return fig.mouse_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    rubberband.mousedown('button_press', mouse_event_fn);\n",
       "    rubberband.mouseup('button_release', mouse_event_fn);\n",
       "    // Throttle sequential mouse events to 1 every 20ms.\n",
       "    rubberband.mousemove('motion_notify', mouse_event_fn);\n",
       "\n",
       "    rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
       "    rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
       "\n",
       "    canvas_div.on(\"wheel\", function (event) {\n",
       "        event = event.originalEvent;\n",
       "        event['data'] = 'scroll'\n",
       "        if (event.deltaY < 0) {\n",
       "            event.step = 1;\n",
       "        } else {\n",
       "            event.step = -1;\n",
       "        }\n",
       "        mouse_event_fn(event);\n",
       "    });\n",
       "\n",
       "    canvas_div.append(canvas);\n",
       "    canvas_div.append(rubberband);\n",
       "\n",
       "    this.rubberband = rubberband;\n",
       "    this.rubberband_canvas = rubberband[0];\n",
       "    this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
       "    this.rubberband_context.strokeStyle = \"#000000\";\n",
       "\n",
       "    this._resize_canvas = function(width, height) {\n",
       "        // Keep the size of the canvas, canvas container, and rubber band\n",
       "        // canvas in synch.\n",
       "        canvas_div.css('width', width)\n",
       "        canvas_div.css('height', height)\n",
       "\n",
       "        canvas.attr('width', width * mpl.ratio);\n",
       "        canvas.attr('height', height * mpl.ratio);\n",
       "        canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
       "\n",
       "        rubberband.attr('width', width);\n",
       "        rubberband.attr('height', height);\n",
       "    }\n",
       "\n",
       "    // Set the figure to an initial 600x600px, this will subsequently be updated\n",
       "    // upon first draw.\n",
       "    this._resize_canvas(600, 600);\n",
       "\n",
       "    // Disable right mouse context menu.\n",
       "    $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
       "        return false;\n",
       "    });\n",
       "\n",
       "    function set_focus () {\n",
       "        canvas.focus();\n",
       "        canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    window.setTimeout(set_focus, 100);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>')\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items) {\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) {\n",
       "            // put a spacer in here.\n",
       "            continue;\n",
       "        }\n",
       "        var button = $('<button/>');\n",
       "        button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
       "                        'ui-button-icon-only');\n",
       "        button.attr('role', 'button');\n",
       "        button.attr('aria-disabled', 'false');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "\n",
       "        var icon_img = $('<span/>');\n",
       "        icon_img.addClass('ui-button-icon-primary ui-icon');\n",
       "        icon_img.addClass(image);\n",
       "        icon_img.addClass('ui-corner-all');\n",
       "\n",
       "        var tooltip_span = $('<span/>');\n",
       "        tooltip_span.addClass('ui-button-text');\n",
       "        tooltip_span.html(tooltip);\n",
       "\n",
       "        button.append(icon_img);\n",
       "        button.append(tooltip_span);\n",
       "\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    var fmt_picker_span = $('<span/>');\n",
       "\n",
       "    var fmt_picker = $('<select/>');\n",
       "    fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
       "    fmt_picker_span.append(fmt_picker);\n",
       "    nav_element.append(fmt_picker_span);\n",
       "    this.format_dropdown = fmt_picker[0];\n",
       "\n",
       "    for (var ind in mpl.extensions) {\n",
       "        var fmt = mpl.extensions[ind];\n",
       "        var option = $(\n",
       "            '<option/>', {selected: fmt === mpl.default_extension}).html(fmt);\n",
       "        fmt_picker.append(option)\n",
       "    }\n",
       "\n",
       "    // Add hover states to the ui-buttons\n",
       "    $( \".ui-button\" ).hover(\n",
       "        function() { $(this).addClass(\"ui-state-hover\");},\n",
       "        function() { $(this).removeClass(\"ui-state-hover\");}\n",
       "    );\n",
       "\n",
       "    var status_bar = $('<span class=\"mpl-message\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
       "    // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
       "    // which will in turn request a refresh of the image.\n",
       "    this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_message = function(type, properties) {\n",
       "    properties['type'] = type;\n",
       "    properties['figure_id'] = this.id;\n",
       "    this.ws.send(JSON.stringify(properties));\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_draw_message = function() {\n",
       "    if (!this.waiting) {\n",
       "        this.waiting = true;\n",
       "        this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
       "    }\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    var format_dropdown = fig.format_dropdown;\n",
       "    var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
       "    fig.ondownload(fig, format);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
       "    var size = msg['size'];\n",
       "    if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
       "        fig._resize_canvas(size[0], size[1]);\n",
       "        fig.send_message(\"refresh\", {});\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
       "    var x0 = msg['x0'] / mpl.ratio;\n",
       "    var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
       "    var x1 = msg['x1'] / mpl.ratio;\n",
       "    var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
       "    x0 = Math.floor(x0) + 0.5;\n",
       "    y0 = Math.floor(y0) + 0.5;\n",
       "    x1 = Math.floor(x1) + 0.5;\n",
       "    y1 = Math.floor(y1) + 0.5;\n",
       "    var min_x = Math.min(x0, x1);\n",
       "    var min_y = Math.min(y0, y1);\n",
       "    var width = Math.abs(x1 - x0);\n",
       "    var height = Math.abs(y1 - y0);\n",
       "\n",
       "    fig.rubberband_context.clearRect(\n",
       "        0, 0, fig.canvas.width, fig.canvas.height);\n",
       "\n",
       "    fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
       "    // Updates the figure title.\n",
       "    fig.header.textContent = msg['label'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
       "    var cursor = msg['cursor'];\n",
       "    switch(cursor)\n",
       "    {\n",
       "    case 0:\n",
       "        cursor = 'pointer';\n",
       "        break;\n",
       "    case 1:\n",
       "        cursor = 'default';\n",
       "        break;\n",
       "    case 2:\n",
       "        cursor = 'crosshair';\n",
       "        break;\n",
       "    case 3:\n",
       "        cursor = 'move';\n",
       "        break;\n",
       "    }\n",
       "    fig.rubberband_canvas.style.cursor = cursor;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_message = function(fig, msg) {\n",
       "    fig.message.textContent = msg['message'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
       "    // Request the server to send over a new figure.\n",
       "    fig.send_draw_message();\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
       "    fig.image_mode = msg['mode'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Called whenever the canvas gets updated.\n",
       "    this.send_message(\"ack\", {});\n",
       "}\n",
       "\n",
       "// A function to construct a web socket function for onmessage handling.\n",
       "// Called in the figure constructor.\n",
       "mpl.figure.prototype._make_on_message_function = function(fig) {\n",
       "    return function socket_on_message(evt) {\n",
       "        if (evt.data instanceof Blob) {\n",
       "            /* FIXME: We get \"Resource interpreted as Image but\n",
       "             * transferred with MIME type text/plain:\" errors on\n",
       "             * Chrome.  But how to set the MIME type?  It doesn't seem\n",
       "             * to be part of the websocket stream */\n",
       "            evt.data.type = \"image/png\";\n",
       "\n",
       "            /* Free the memory for the previous frames */\n",
       "            if (fig.imageObj.src) {\n",
       "                (window.URL || window.webkitURL).revokeObjectURL(\n",
       "                    fig.imageObj.src);\n",
       "            }\n",
       "\n",
       "            fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
       "                evt.data);\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "        else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
       "            fig.imageObj.src = evt.data;\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        var msg = JSON.parse(evt.data);\n",
       "        var msg_type = msg['type'];\n",
       "\n",
       "        // Call the  \"handle_{type}\" callback, which takes\n",
       "        // the figure and JSON message as its only arguments.\n",
       "        try {\n",
       "            var callback = fig[\"handle_\" + msg_type];\n",
       "        } catch (e) {\n",
       "            console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        if (callback) {\n",
       "            try {\n",
       "                // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
       "                callback(fig, msg);\n",
       "            } catch (e) {\n",
       "                console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
       "            }\n",
       "        }\n",
       "    };\n",
       "}\n",
       "\n",
       "// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
       "mpl.findpos = function(e) {\n",
       "    //this section is from http://www.quirksmode.org/js/events_properties.html\n",
       "    var targ;\n",
       "    if (!e)\n",
       "        e = window.event;\n",
       "    if (e.target)\n",
       "        targ = e.target;\n",
       "    else if (e.srcElement)\n",
       "        targ = e.srcElement;\n",
       "    if (targ.nodeType == 3) // defeat Safari bug\n",
       "        targ = targ.parentNode;\n",
       "\n",
       "    // jQuery normalizes the pageX and pageY\n",
       "    // pageX,Y are the mouse positions relative to the document\n",
       "    // offset() returns the position of the element relative to the document\n",
       "    var x = e.pageX - $(targ).offset().left;\n",
       "    var y = e.pageY - $(targ).offset().top;\n",
       "\n",
       "    return {\"x\": x, \"y\": y};\n",
       "};\n",
       "\n",
       "/*\n",
       " * return a copy of an object with only non-object keys\n",
       " * we need this to avoid circular references\n",
       " * http://stackoverflow.com/a/24161582/3208463\n",
       " */\n",
       "function simpleKeys (original) {\n",
       "  return Object.keys(original).reduce(function (obj, key) {\n",
       "    if (typeof original[key] !== 'object')\n",
       "        obj[key] = original[key]\n",
       "    return obj;\n",
       "  }, {});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.mouse_event = function(event, name) {\n",
       "    var canvas_pos = mpl.findpos(event)\n",
       "\n",
       "    if (name === 'button_press')\n",
       "    {\n",
       "        this.canvas.focus();\n",
       "        this.canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    var x = canvas_pos.x * mpl.ratio;\n",
       "    var y = canvas_pos.y * mpl.ratio;\n",
       "\n",
       "    this.send_message(name, {x: x, y: y, button: event.button,\n",
       "                             step: event.step,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "\n",
       "    /* This prevents the web browser from automatically changing to\n",
       "     * the text insertion cursor when the button is pressed.  We want\n",
       "     * to control all of the cursor setting manually through the\n",
       "     * 'cursor' event from matplotlib */\n",
       "    event.preventDefault();\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    // Handle any extra behaviour associated with a key event\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.key_event = function(event, name) {\n",
       "\n",
       "    // Prevent repeat events\n",
       "    if (name == 'key_press')\n",
       "    {\n",
       "        if (event.which === this._key)\n",
       "            return;\n",
       "        else\n",
       "            this._key = event.which;\n",
       "    }\n",
       "    if (name == 'key_release')\n",
       "        this._key = null;\n",
       "\n",
       "    var value = '';\n",
       "    if (event.ctrlKey && event.which != 17)\n",
       "        value += \"ctrl+\";\n",
       "    if (event.altKey && event.which != 18)\n",
       "        value += \"alt+\";\n",
       "    if (event.shiftKey && event.which != 16)\n",
       "        value += \"shift+\";\n",
       "\n",
       "    value += 'k';\n",
       "    value += event.which.toString();\n",
       "\n",
       "    this._key_event_extra(event, name);\n",
       "\n",
       "    this.send_message(name, {key: value,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
       "    if (name == 'download') {\n",
       "        this.handle_save(this, null);\n",
       "    } else {\n",
       "        this.send_message(\"toolbar_button\", {name: name});\n",
       "    }\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
       "    this.message.textContent = tooltip;\n",
       "};\n",
       "mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
       "\n",
       "mpl.extensions = [\"eps\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\"];\n",
       "\n",
       "mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
       "    // Create a \"websocket\"-like object which calls the given IPython comm\n",
       "    // object with the appropriate methods. Currently this is a non binary\n",
       "    // socket, so there is still some room for performance tuning.\n",
       "    var ws = {};\n",
       "\n",
       "    ws.close = function() {\n",
       "        comm.close()\n",
       "    };\n",
       "    ws.send = function(m) {\n",
       "        //console.log('sending', m);\n",
       "        comm.send(m);\n",
       "    };\n",
       "    // Register the callback with on_msg.\n",
       "    comm.on_msg(function(msg) {\n",
       "        //console.log('receiving', msg['content']['data'], msg);\n",
       "        // Pass the mpl event to the overridden (by mpl) onmessage function.\n",
       "        ws.onmessage(msg['content']['data'])\n",
       "    });\n",
       "    return ws;\n",
       "}\n",
       "\n",
       "mpl.mpl_figure_comm = function(comm, msg) {\n",
       "    // This is the function which gets called when the mpl process\n",
       "    // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
       "\n",
       "    var id = msg.content.data.id;\n",
       "    // Get hold of the div created by the display call when the Comm\n",
       "    // socket was opened in Python.\n",
       "    var element = $(\"#\" + id);\n",
       "    var ws_proxy = comm_websocket_adapter(comm)\n",
       "\n",
       "    function ondownload(figure, format) {\n",
       "        window.open(figure.imageObj.src);\n",
       "    }\n",
       "\n",
       "    var fig = new mpl.figure(id, ws_proxy,\n",
       "                           ondownload,\n",
       "                           element.get(0));\n",
       "\n",
       "    // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
       "    // web socket which is closed, not our websocket->open comm proxy.\n",
       "    ws_proxy.onopen();\n",
       "\n",
       "    fig.parent_element = element.get(0);\n",
       "    fig.cell_info = mpl.find_output_cell(\"<div id='\" + id + \"'></div>\");\n",
       "    if (!fig.cell_info) {\n",
       "        console.error(\"Failed to find cell for figure\", id, fig);\n",
       "        return;\n",
       "    }\n",
       "\n",
       "    var output_index = fig.cell_info[2]\n",
       "    var cell = fig.cell_info[0];\n",
       "\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.handle_close = function(fig, msg) {\n",
       "    var width = fig.canvas.width/mpl.ratio\n",
       "    fig.root.unbind('remove')\n",
       "\n",
       "    // Update the output cell to use the data from the current canvas.\n",
       "    fig.push_to_output();\n",
       "    var dataURL = fig.canvas.toDataURL();\n",
       "    // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
       "    // the notebook keyboard shortcuts fail.\n",
       "    IPython.keyboard_manager.enable()\n",
       "    $(fig.parent_element).html('<img src=\"' + dataURL + '\" width=\"' + width + '\">');\n",
       "    fig.close_ws(fig, msg);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.close_ws = function(fig, msg){\n",
       "    fig.send_message('closing', msg);\n",
       "    // fig.ws.close()\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
       "    // Turn the data on the canvas into data in the output cell.\n",
       "    var width = this.canvas.width/mpl.ratio\n",
       "    var dataURL = this.canvas.toDataURL();\n",
       "    this.cell_info[1]['text/html'] = '<img src=\"' + dataURL + '\" width=\"' + width + '\">';\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Tell IPython that the notebook contents must change.\n",
       "    IPython.notebook.set_dirty(true);\n",
       "    this.send_message(\"ack\", {});\n",
       "    var fig = this;\n",
       "    // Wait a second, then push the new image to the DOM so\n",
       "    // that it is saved nicely (might be nice to debounce this).\n",
       "    setTimeout(function () { fig.push_to_output() }, 1000);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>')\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items){\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) { continue; };\n",
       "\n",
       "        var button = $('<button class=\"btn btn-default\" href=\"#\" title=\"' + name + '\"><i class=\"fa ' + image + ' fa-lg\"></i></button>');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    // Add the status bar.\n",
       "    var status_bar = $('<span class=\"mpl-message\" style=\"text-align:right; float: right;\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "\n",
       "    // Add the close button to the window.\n",
       "    var buttongrp = $('<div class=\"btn-group inline pull-right\"></div>');\n",
       "    var button = $('<button class=\"btn btn-mini btn-primary\" href=\"#\" title=\"Stop Interaction\"><i class=\"fa fa-power-off icon-remove icon-large\"></i></button>');\n",
       "    button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
       "    button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
       "    buttongrp.append(button);\n",
       "    var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
       "    titlebar.prepend(buttongrp);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(el){\n",
       "    var fig = this\n",
       "    el.on(\"remove\", function(){\n",
       "\tfig.close_ws(fig, {});\n",
       "    });\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(el){\n",
       "    // this is important to make the div 'focusable\n",
       "    el.attr('tabindex', 0)\n",
       "    // reach out to IPython and tell the keyboard manager to turn it's self\n",
       "    // off when our div gets focus\n",
       "\n",
       "    // location in version 3\n",
       "    if (IPython.notebook.keyboard_manager) {\n",
       "        IPython.notebook.keyboard_manager.register_events(el);\n",
       "    }\n",
       "    else {\n",
       "        // location in version 2\n",
       "        IPython.keyboard_manager.register_events(el);\n",
       "    }\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    var manager = IPython.notebook.keyboard_manager;\n",
       "    if (!manager)\n",
       "        manager = IPython.keyboard_manager;\n",
       "\n",
       "    // Check for shift+enter\n",
       "    if (event.shiftKey && event.which == 13) {\n",
       "        this.canvas_div.blur();\n",
       "        event.shiftKey = false;\n",
       "        // Send a \"J\" for go to next cell\n",
       "        event.which = 74;\n",
       "        event.keyCode = 74;\n",
       "        manager.command_mode();\n",
       "        manager.handle_keydown(event);\n",
       "    }\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    fig.ondownload(fig, null);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.find_output_cell = function(html_output) {\n",
       "    // Return the cell and output element which can be found *uniquely* in the notebook.\n",
       "    // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
       "    // IPython event is triggered only after the cells have been serialised, which for\n",
       "    // our purposes (turning an active figure into a static one), is too late.\n",
       "    var cells = IPython.notebook.get_cells();\n",
       "    var ncells = cells.length;\n",
       "    for (var i=0; i<ncells; i++) {\n",
       "        var cell = cells[i];\n",
       "        if (cell.cell_type === 'code'){\n",
       "            for (var j=0; j<cell.output_area.outputs.length; j++) {\n",
       "                var data = cell.output_area.outputs[j];\n",
       "                if (data.data) {\n",
       "                    // IPython >= 3 moved mimebundle to data attribute of output\n",
       "                    data = data.data;\n",
       "                }\n",
       "                if (data['text/html'] == html_output) {\n",
       "                    return [cell, data, j];\n",
       "                }\n",
       "            }\n",
       "        }\n",
       "    }\n",
       "}\n",
       "\n",
       "// Register the function which deals with the matplotlib target/channel.\n",
       "// The kernel may be null if the page has been refreshed.\n",
       "if (IPython.notebook.kernel != null) {\n",
       "    IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
       "}\n"
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<img src=\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAArwAAAK8CAYAAAANumxDAAAgAElEQVR4Xuy9CbhXZbn3fyNgBDI44JSadhQRBxSPSmllZFaOOKT2OgBqZopTXg45Vcc0tByQY9RfQ0jU4/CaWna9ZjmUqeCs4YDFK4ppYCqCKJP8r3v1btwb9vD93Xtt9t7r/qzrOpfhvp/1W8/3/rCfz1mu37O6LFu2bJlxkAAJkAAJkAAJkAAJkEBFE+iC8Fa0s0yLBEiABEiABEiABEigSADhBQQSIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwVrq9TI4ESIAESIAESIAESADhhQESIAESIAESIAESIIFKJ4DwtmF7u3TpUpx99dVXb8NP4dQkQAIkQAIkQAIkYLZo0aIihmXLlhHHCgkgvG2IBML7cbgfffRR8YfVVlutDRPn1J01AfjorJ1bNdcNH6sm5874KbDRsGsIb9MUI7xt+Df8E5/4RHH2hQsXtuGndI5T/+Y3vykudN999+0cF8xVrtIE4GOVxt3pPgw+Ol3LVtkFw0bDqPEOhHeV/eWr/0GA93Ea/FJqFwQ7zYfCR6dpVbtcKHy0S+yd4kNhA+FVQeUOr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0pQrvvHnz7P7777fHH3+8+L8nnnjCZs+eXVyL//vdd99dva5G6xYvXmzjxo2zG264wV5++eWiZsCAAXb44Yfb6NGjrXv37s2e/+mnn7af/OQn9sADD9hbb71l/fv3L67pjDPOsMGDB7fq2hobjPAivKVDVdETsmhVtLElTQs+SgqygqeBDYRXxbpU4b3jjjvsgAMOaPSzWyu88+fPtz322MOmTJlSnL9Hjx7FPz/88MPin5/97Gft3nvvtV69ejX6+TfeeKONHDnSXJr96Nu3r82dO7f436uvvrpNmjTJDjvsMDU3qQ7hRXglUCgyFi0gaC4B+ICPphKADYRX/dtRuvAed9xxtuOOO9p//ud/2qBBg+x//a//VVxLa4XX7+K6tPbr188mTJhgw4cPL87rkn300Ufbu+++a0cddVQhrise06ZNsyFDhtiiRYvskEMOsSuvvNI22GADe+ONN+yUU06xW2+91VxO/Q7wwIED1exarEN4Ed4WIaGgSIBFCxAQXhiIJMDvDoRX5aZU4V26dKl17dp1+Wf7XdnevXu3WnifffZZ23777W3ZsmV222232UEHHdRgfv7vvvGNb1iXLl3Ma7fZZpsGP/f622+/vTjHY489Zt26dVv+8yVLlthOO+1UyK6f45ZbblGza7EO4UV4W4SEAoQXBlpMAKlpMaK0BbCB8Krwlyq8K35oWcJ71lln2aWXXmqbb765TZ8+vRDb+oeLsD/L+7e//c2+973v2cUXX7z8x/7Ywrrrrlvc3b3++uvtiCOOWCmbyZMn25FHHlnc5Z0zZ85ySVdDbKoO4UV4W8tQlvEsWlk6HZsnfMRyyzAKNhBelfNOIbx+B9a/BHfiiSfaf//3fzc6N//Zz372M9t5552XP+frhXfffbfts88+xRj/Ap1/UW3FwyXXpdiP3/3ud/b1r39dza/ZOoQX4S0FpAQnYdFK0ORWTBE+WhFexYfCBsKrIt7hhdfv3vpjEe+//76NHz/ejj/++Ebn5j874YQTitr33ntveY3vynDmmWfaeuutZ2+++WaTufjPXYh/+tOf2umnn67mh/CKSfFLSQwqaRl8JG28OG34EINKWAYbCK+KfYcXXpdX31HBjzvvvNP222+/RufmP6v7Iptvj7bGGmsUdd/97nftiiuusB122MGefPLJJnPxn/tzvC67Lr1lHNzh5Q5vGRxlOAeLVoYux+cIH/HsVsXIme+8bTc8OTX0UfPmvmt/e+6J0FgfNGfOv7c+7d//3/+VNnJsvu2O1rtvv8hQG7rZ2jZ0462tZ/eeofFlD8I7mk60wwvvP/7xD/vUpz5VzMC3HfOtyRo7/Gd77rln8SMf47sw+OG7RlxzzTW266672kMPPdRkEv7zhx9+uKj/xS9+ITFYB1ZTxf7csH9Bzr8wl/346KOPighWW2217FEw/0YSgA+waC4B+OjYfNzxxms29RPjO/ZFtuHVHd7jO7Z1z43b8BP0Ux944IHFOrtw4UJ9UJJKhPf/NRrhbVviWbDaNt/Ofnb46OwdbNvrh4+2zbe1Z0d4Ed7WMrQqxnd44eWRhlWBQdt/Bv9Jsu0z7syfAB+duXttf+3w0fYZt+YTeKSBRxpaw8+qGtvhhde/tNanTx/zLc4iX1rz53H91cF8aW1VIdX457BgtW/+Hf3T4aOjd6h9rw8+2jf/jvzpsNGwOzzD2zStHV54/dJ9qzF/YcTo0aNt3Lhxjc7Gf3b11VevtC2ZbzO29957F2N8+7F11llnpfFsS9b2v874pdT2GXfmT4CPzty9tr92+Gj7jDvrJ8AGwquy2ymE9+yzz7ZLLrnEtthii+LFE40d/uKJl19+eaUXT/gjEb73rn+BzF8w4a8oXvG44YYbihdS8OIJFZva6/ilVHtmmUbAR6Zu1z5X+Kg9sywjYAPhVVnvFMJb/9XCvuPBAQcc0GB+/u/89cEtvVp4yJAhNnXq1AavP/ZXC++yyy7FlmW8WljFpvY6finVnlmmEfCRqdu1zxU+as8sywjYQHhV1ksX3rfeemv5Z/vLIjbddNPiz3fccUexNVjd4Xvrdu/effmfvW7mzJk2YsQImzhx4krX73dmb7zxRltzzTXtuuuuW74f71133WWjRo2yd955x4466iibNGnSSmOnTZtmLrt+l/fQQw+1K6+80tZff/3iRRSnnnqq3XzzzcXdXd+Hd+DAgWp2LdbxLM3HEfFLqUVcUhfAR+r2tzh5+GgxorQFsIHwqvCXLrx+l1U57r//ftt9991l4fUvrfkevFOmTCnGfPKTnzT/QtuHH35Y/Pmzn/1ssU9vr169Gv14l+WRI0fa4sWLizvBLtxz584tzuHi/atf/coOO+ww5dLlGoQX4ZVhSV7IopUcgBamDx/w0VQCsIHwqn87Oo3w+oRcVv1La/7Mbd2zvP7srj9/619aq3/HuLEA/A7upZdeag8++KD5nWj/AptLt796ePDgwWpmch3Ci/DKsCQvZNFKDgDCCwDBBPjdgfCq6JQuvOoHZ6hDeBHeDJyXMUcWrTJSrO454KO6vW3tzGAD4VUZQnjVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhR3jVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvApl2GzHznbbvhyanhz543913723NPhMfPmTO7GNu//7rhc2y+7Y7Wu2+/mscP3WxtG7rx1taze8+axzJg1SSA1KyanDvjp8AGwqtyi/CqSQXqEF6EN4BNuwy5+I//x26adUa7fHZH+NCxn59gwz6zU0e4FK6hkQSQGrBoKgHYQHjVvx0Ir5pUoA7hRXgD2LTLEIQX4W0X8MQPRWrEoBKWwQbCq2KP8KpJBeoQXoQ3gE27DOGRBh5paBfwxA9FasSgEpbBBsKrYo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQt8mwjtnzhwbM2aM3XXXXTZr1izr1auXDRkyxE444QQbPny4em0r1T322GN2xRVX2J///GebPXu29e3btzjvcccdZwceeGCT5x05cqRNmjSp2c/deuut7a9//Wv42hobiPAivKUCVeGTsWhVuLklTA0+SgixoqeADYRXRbt04Z02bZoNGzasEFI/evfubQsWLLClS5cWfz755JNt7Nix6vUtr7vsssvszDPPtI8++si6dOli/fr1s3nz5tmSJUuKmqOPPtquvfba4mcrHnXC26NHj0KSGzsGDhxoDzzwQM3X1dwAhBfhLRWoCp+MRavCzS1havBRQogVPQVsILwq2qUK78KFC23QoEE2Y8YM22abbWzy5Mk2ePDgQnj9zuz5559vy5YtswkTJtioUaPUa7Tf/va3tu+++xb1RxxxhF166aW2wQYb2AcffGDXXXednXbaabZo0SK75JJLCiluSnhHjBhhEydOlD+3tYUIL8LbWoayjGfRytLp2DzhI5ZbhlGwgfCqnJcqvOPGjSvu4Pbs2dNeeOEF22STTRpcx+jRo+3qq6+2DTfc0F555RXr3r27dJ3+2MJTTz1lO+20k02ZMmWlu7gXXnihXXDBBcXd5JkzZ9qaa67Z4Lx1d3gRXinuNinil1KbxFqZk8JHZVrZJhOBjzaJtRInhQ2EVwW5VOHdZZddbOrUqXbMMccUjxeseLz66qu26aabFnd577nnHttzzz1bvM4333yzuJvrx0033WSHHXbYSmPef/9969OnT/G4wy9/+cvi8Yb6B8LbYsxtXsAvpTaPuFN/AHx06va1+cXDR5tH3Gk/ADYQXhXe0oR3/vz5hXS6zN5666128MEHN3oN/uWw559/vnj0wB9BaOnwO7pDhw4tyvwu7/bbb9/oEBdpv7vrQuxijPC2lOyq/Tm/lFZt3p3t0+Cjs3Vs1V4vfKzavDvTp8EGwqvyWprw+g4KO++8c/G5/jiDfwmssePQQw+1W265xfbZZx+rA7W5i/U7xn7n2I/HH3/cdtxxx0bLN9poI3v99ddt2223tWeffbZR4fUa/+Ka32n2f26++ea21157mT9qsd5666mZyXU8w/txVPxSkrFJWQgfKdsuTxo+5KjSFcIGwqtCX5rw+hZk+++/f/G5c+fOLe72Nnb4F8yuvPLKQlxdYFs6/vnPf9r6669flPnWYkcdddRKQ959911ba621irvLa6+9tr311luNCq//y65duxbX5tfoj0D44WNdwr/85S+3dDkNfl4ntE0N8i/SdevWzW6//faazlvF4rqsV1tttSpOjzm1MgH4aGWAFR8OHxVvcCumBxsNw/MtWn2d9U0EOBomUJrw3njjjXb44YcXZ1+8eHEheo0d5557rl188cU2YMAAe+mll6R+7LDDDvb000+bPw7xzDPPFNJa/zj77LOXPx6x+uqrr9Toq666qgDA7+Z++tOfLsb7lmYu6WeccYa98cYbtsYaa9gTTzxRXJd6ILxqUrb8/7lAePXMMlWyaGXqdu1zhY/aM8syAjYQXpX1TiG8d9555/IXVri0ujD79me+1+8111xjvkuDS6yLtj+q4NuVqYfvFuG7QLzzzjv2zW9+01zcyzp4pOHjJPnPTmVRVc3zwEc1+1rWrOCjrCSrdx7YaNhTvKNpxksT3rZ6pKHu0n3vXb+T648trHjUPYvrd3L98Qe/Y1vL4fsD/+hHPyoedXDxLesuJOAhvLVwmLmWRStz91ueO3y0nFHWCthAeFX2SxNefx7X98n1o8wvrdWfiO/SMH78eHv00UfNn9t1ufUXUvhzwd/97neLu71f+MIX7MEHH1TnX9TdfffdxZfo/PC7xv37969pfFPFCC/CWwpICU7CopWgya2YIny0IryKD4UNhFdFvDThrb8t2W233WYHHXRQo9fgb2Dz1w+r25KpE9lqq63sxRdftLPOOsvGjBmjDltJeOfMmWPrrLNOTeMR3pbj4pdSyxllroCPzN1vee7w0XJGWStgA+FV2S9NeP0Dfb9c3zf32GOPLe62rnjMmjWrePtaLS+eUCbyyCOP2Oc+97mi9Lnnnitea1zL4W9p8+eA/U1tfueYRxpqSU+r5ZeSllPWKvjI2nlt3vCh5ZSxCjYQXpX7UoW37tXCvXr1Kh5r2HjjjRtch7922GtqfbVwc5Pxt6x9/vOfL15Ksd9++5l/wa3+4XLdpUuXJk/he/L6LhBvv/12oy+tUINsrI5HGj5OhV9KrSGp+mPho/o9bs0M4aM16VV7LGwgvCrhpQqv7/vmuyfMmDHDtttuO7v++uuLf/quCWPHjrVzzjmnuLs7YcIEGzVqVINrrHtT2ogRI2zixIkNfuZ78V5++eV2yCGHFFuT+U4MS5Yssfvuu694hMG3LPPneX1bMZfp+odfwx133GFHHnmk7bbbbssfV/BHMPwvim9L5i+scEn38VtuuaWaXYt1CC/C2yIkFBQJsGgBQnMJwAd8NJUAbCC86t+OUoXXP9Sfzx02bFjx5S8/fOeDBQsWFILqx0knnWS+m8KKR3PC61uHbbbZZsUQv1vbr1+/Yh/dunNuscUWxYLZmKy6PNeXa99v10XUd2Oo27/Pv6R2880325e+9CU1N6kO4UV4JVAoQnhhoNkEkBoAQXg1BvCOpnMqXXj9o1x2/Ytj/kvqtddeK+6e+l63J5544vL9dGsRXn9s4bLLLrP777/fpk+fXrxJzUXav6j2jW98w4477rhCYhs7Zs6caZMnT7a//OUvxYsufKwLeN++fYu70Xvvvbd961vfKt62VvYBeAhv2UxV9XwITVU7W8684KOcHKt4Ftho2FW8YxULbxX/UkXmBHgIb4SbjGNYtDJ2XZ8zfOhZZauEDYRXZb5N7vCqH171OoQX4a0642XNj0WrrCSreR74qGZfy5gVbCC8KkcIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHM/YdwMAACAASURBVMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoK5Kwjt//nybNm1aIAWzDxYtsd/c/0gxdt8vfdY+uXq30Hl80NZbb21rrLFGeDwDO2YCLFodsy8d5argo6N0ouNdB2wgvCqVCK+aVKCuIwrvex8utr/Pnl/zbJ576nEbMXzPmseVPWDSHb+3bXf4z5pP+x/rrmF9enSveRwDVk0CLFqrJufO+inw0Vk71/bXDRsIr0oZwqsmFajraML7f//1tn153M2BmZgtfPNv9s6940Njyxy05le+Y59Yf/OaT9nrE13t3tGH2vq9+9Y8lgFtnwCLVttn3Jk/AT46c/fa9tphA+FVCUN41aQCdR1NeC/+4/+xm2adEZhJNYZ8f8jP7eBtd63GZCo2CxatijW05OnAR8mBVuh0sIHwqjgjvGpSgTqENxBaGw75//aYZJ/91JA2/AROHU2ARSuaXI5x8JGjz5FZwgbCq3KD8KpJBeo6mvDOfOdtu+HJqcVMNlmrp/Xo1lWe1YcfLrBZr86U61csfPnll4t/tcUWW4TP4QM32uTT1qNHz5rP8am1PmmD1x1oPbvXPrbmD2NAzQmwaNUcWaoB8JGq3TVNFjYQXhWYNhHeOXPm2JgxY+yuu+6yWbNmWa9evWzIkCF2wgkn2PDhw9VrW6nusccesyuuuML+/Oc/2+zZs61v377FeY877jg78MADWzzv008/bT/5yU/sgQcesLfeesv69+9vu+++u51xxhk2ePDgFsfXWtDRhLfW6y+znl9KZaZZvXPBR/V6WuaM4KPMNKt1LthAeFWiSxde37pq2LBhhZD60bt3b1uwYIEtXbq0+PPJJ59sY8eOVa9ved1ll11mZ555pn300UfWpUsX69evn82bN8+WLFlS1Bx99NF27bXXFj9r7Ljxxhtt5MiRtnjx4uLHLstz584t/vfqq69ukyZNssMOO6zm62puAML7cTr8UioVrcqdDD4q19JSJwQfpcZZqZPBBsKrAl2q8C5cuNAGDRpkM2bMsG222cYmT55c3Dl14fU7s+eff74tW7bMJkyYYKNGjVKv0X7729/avvvuW9QfccQRdumll9oGG2xgH3zwgV133XV22mmn2aJFi+ySSy4ppHjFwyXc7wR7zSGHHGJXXnllMf6NN96wU045xW699VZzOfU7wAMHDpSvq6VChBfhbYkRfv7vBFi0IKG5BOADPppKADYQXvVvR6nCO27cuOIObs+ePe2FF16wTTbZpMF1jB492q6++mrbcMMN7ZVXXrHu3bV9UV1Wn3rqKdtpp51sypQpK93FvfDCC+2CCy4o7ibPnDnT1lxzzQafe9BBB9ntt99u22+/vfljEd26ffziA79D7Od12f3GN75ht9xyi5pdi3UIL8LbIiQUILww0GICSE2LEaUtgA2EV4W/VOHdZZddbOrUqXbMMccUjxeseLz66qu26aabFnd577nnHttzz5ZfZPDmm28Wd2P9uOmmmxp97OD999+3Pn36FI87/PKXvyweb6g7/LGFddddt7i7e/311xd3iFc8/E70kUceWdzl9eePXZzLOBBehLcMjjKcg0UrQ5fjc4SPeHZVHwkbCK/KeGnC66+edel0mfVHBA4++OBGr8FfDfv8888Xjx74IwgtHX5Hd+jQoUWZ3+X1u7SNHS7SfnfXn8N1Ma477r77bttnn32KP/pzxf5FtRUPl1yXYj9+97vf2de//vWWLkv6OcKL8EqgUMQjDTDQbAJIDYA0lQBsILzq347ShNcfFdh5552Lz/XHGZp6FvbQQw8tHhtwCa0DtbmL9TvGfufYj8cff9x23HHHRss32mgje/31123bbbe1Z599dnmN78rgcr3eeuuZ3y1u6vCfuxD/9Kc/tdNPP13Nr9k6hBfhLQWkBCdh0UrQ5FZMET5aEV7Fh8IGwqsiXprw+hZk+++/f/G5/hiB3+1t7PAvmPmXxlxcXWBbOv75z3/a+uuvX5T5TgpHHXXUSkPeffddW2uttYq7y2uvvXax5Vjd8d3vfrf4wtwOO+xgTz75ZJMf5z/353hddl16yzgQXoS3DI4ynINFK0OX43OEj3h2VR8JGwivynhpwuvbfh1++OHF5/rWX/W/GFb/Ys4991y7+OKLbcCAAfbSSy9J11kno/44xDPPPGNduzZ8YcLZZ5+9/PEI32LMd4uoO3yP3muuucZ23XVXe+ihh5r8PP/5ww8/XOzp+4tf/EK6rjqhbarYnxv2HPwLc9kPf77aj9VWWy17FMy/kQTgAyyaSwA+4KOpBGCjYTL+TgJfZ+t7EPT8O4FOIbx33nnn8hdW7LXXXoUw+/Zn/giCy6zv0uAS7KLdo0ePYrsyhLdjIc4vpY7Vj452NfDR0TrSsa4HPjpWPzrS1cAGwqvyWJrwttUjDXUT8b13/U6uP7aw4rH55pubi/BVV11VPP7g++vWHTzSoKLQtnX8Z6e2zbeznx0+OnsH2/b64aNt8+3MZ4eNht3jUcqmaS5NeP15XN/P1o8yv7RW/9J9l4bx48fbo48+av7crsutv5DCnwt2sfW7vV/4whfswQcfXD7Mn8f1VwfzpbX2/ZXGL6X2zb+jfzp8dPQOte/1wUf75t+RPx02EF6Vz9KEt/62ZLfddpv5yx4aO/wNbP7mM3VbMnUiW221lb344ot21lln2ZgxY5YP823G9t577+LPvv3YOuuss9Ip2ZZMTTlexy+leHYZRsJHhi7H5wgf8eyqPhI2EF6V8dKE1z/Q98v1fXOPPfbY4m7risesWbOKt6/V8uIJZSKPPPKIfe5znytKn3vuueK1xnXHe++9V+y9618g8xdM1H2xrv55b7jhhuKFFLx4Qkk7VsMvpVhuWUbBR5ZOx+YJH7HcMoyCDYRX5bxU4a17tXCvXr2Kxxo23njjBtfhrx32mlpfLdzcZPwta5///OeLl1Lst99+5l9wW/Goe7Wwv6LY9/Wtv8uDv1rY9/n1Lct4tbCKTe11/FKqPbNMI+AjU7drnyt81J5ZlhGwgfCqrJcqvL4Nhu+eMGPGDNtuu+2KV/n6P33XhLFjx9o555xT3N2dMGGCjRo1qsE11r0pbcSIETZx4sQGP/O9eC+//HI75JBDzLcm850YXFTvu+++4hEG3z/Xn+d94oknCple8fBHKFx2/S6vv/jC9wH2en8Rxamnnmo333xzcXfXz9PUCzPUQOvX8fD4x2nwSylCUJ4x8JGn15GZwkcktRxjYAPhVUkvVXj9Q10uhw0bVmwZ5oe/gGLBggWFoPpx0kknFbsprHg0J7yvvPKKbbbZZsWQLl26WL9+/WzevHnLz7nFFlsUb23bcsstm5y37xM8cuTIYusyP0ffvn2LF2S4gHfv3t1+9atfFa8lLvNAeBHeMnmq8rlYtKrc3dbPDT5an2FVzwAbCK/KdunC6x/ssutfHHMQX3vtNfNHHPwO64knnrh8P91ahNcfW7jsssvs/vvvt+nTpxdvUnOR9i+q+WMI/rKIll4C4Z/nd3B9ezPfxcHP4V9g23333Ysv0A0ePFjNTK5DeBFeGZbkhSxayQFoYfrwAR9NJQAbCK/6t6NNhFf98KrXIbwIb9UZL2t+LFplJVnN88BHNftaxqxgA+FVOUJ41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhR3jVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhR3jVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhR3jVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhR3jVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhR3jVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhR3jVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhR3jVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhR3jVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhR3jVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhR3jVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhR3jVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhR3jVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhR3jVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhR3jVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhR3jVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhR3jVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhR3jVpAJ1CC/CG8Am5RAWrZRtlycNH3JU6QphA+FVoUd41aQCdQgvwhvAJuUQFq2UbZcnDR9yVOkKYQPhVaFHeNWkAnUIL8IbwCblEBatlG2XJw0fclTpCmED4VWhbxPhnTNnjo0ZM8buuusumzVrlvXq1cuGDBliJ5xwgg0fPly9tpXq/vSnP9n48ePt4YcftjfffNO6du1qG220kX3hC1+wk046yQYPHtzouXfffXd78MEHm/3cvffe237729+Gr62xgQgvwlsqUBU+GYtWhZtbwtTgo4QQK3oK2EB4VbRLF95p06bZsGHDbPbs2cU19O7d2xYsWGBLly4t/nzyySfb2LFj1etbXnfeeefZRRddtPzPPXv2LM65cOHC4t+5/F599dX27W9/e6Vz1wmvi/caa6zR6Gd/5Stfseuvv77m62puAMKL8JYKVIVPxqJV4eaWMDX4KCHEip4CNhBeFe1Shdflc9CgQTZjxgzbZpttbPLkycVdVxfeK664ws4//3xbtmyZTZgwwUaNGqVeo91777225557FvWHHHKI/fjHP7bPfOYzxbmeeeYZO+WUU8zv/nbr1s1cuAcMGNDg3HXC+/3vf99+8IMfyJ/b2kKEF+FtLUNZxrNoZel0bJ7wEcstwyjYQHhVzksV3nHjxhV3cP3u6wsvvGCbbLJJg+sYPXp0cRd2ww03tFdeecW6d+8uXefIkSNt0qRJtvnmmxfndbGtf8ydO7d4tGH+/Pl26aWX2hlnnIHwSsmuuiJ+Ka26rDvjJ8FHZ+zaqrtm+Fh1WXe2T4INhFdltlTh3WWXXWzq1Kl2zDHH2LXXXrvSNbz66qu26aabFndm77nnnuV3bVu62K997WtF/UEHHWS33XZbo+U77rijPfnkk8UdXL+TW//gDm9LCbf9z/ml1PYZd+ZPgI/O3L22v3b4aPuMO+snwAbCq7JbmvD63dU+ffoUMnvrrbfawQcf3Og1bL311vb888/bmWeeaZdccol0nd/5znfs5z//uW2xxRbF2Mbu8G688cY2b9684oty++67L8IrJbvqiviltOqy7oyfBB+dsWur7prhY9Vl3dk+CTYQXpXZ0oT3scces5133rn4XH/sYODAgY1ew6GHHmq33HKL7bPPPlYHaksX+9RTT5nfwXWZ9vEXX3zx8md4n3322eIxCn+G178s94c//MG6dOnSqPD6IxGLFi2yN954o/jy2lZbbWX777+/HX/88YWsl33wDO/HifJLqWy6qnU++KhWP8ueDXyUnWh1zgcbCK9Kc2nC63dWXR798GdqmxLI0047za688spCYB9//HH1OosvwB133HH2wQcfFGPq79Kw/vrr29FHH20XXHCB1Ulm/RPX35Zs9dVXL8b6NbpA++HPGvv1N7WtmXyRKxQivAhvlJ1s41i0snW8tvnCR215ZaqGDYRX5b004b3xxhvt8MMPLz538eLFKz12UHdB5557bnGH1ndSeOmll9TrLOruu+8+O+KII4o7tPWPfv362ZFHHmkXXnih9e3bd6Vz+jO9/uywbz3mX5hbbbXV7O2337abbrrJzjnnHHvvvfdsgw02sOeee87WXntt+Zoak+v6g/1usj9+cfvtt8vnrGrhRx99VEzNs+cggRUTgA+YaC4B+ICPphKAjYbJHHjggcU6W7dlK+R8nECnEN4lS5YUL5bw53j9sQkXZn+RhQvllClTiueBXZ59K7SHHnqoUeltqul+l/lzn/tcIenf+973inOrB8KrJmXGLyU9q4yV8JGx6/qc4UPPKlslbCC8KvOlCW9bPtLgb21zGfXngv153h49ejSYn7/Zzb8M5/88++yzi316azn87rA/MuHC7Hd5yzp4pOHjJPnPTmVRVc3zwEc1+1rWrOCjrCSrdx7YaNhTvKNpxksTXr9TutNOOxWfVPaX1tZbb73izW2+z6/v5dvYcfrpp9vll19uW265pb344os1/a32vYH9vP5s7/vvv1/T2OaKAQ/hLQ2mip+IRaviDW7l9OCjlQFWeDhsILwq3qUJb/1tyXyvXN8zt7HD76L629DUbcn+9a9/2TrrrFOc6u6777a99tqr0fPWSatL5ocffqjOv6irG+uvHvZ5lHUgvAhvWSxV/TwsWlXvcOvmBx+ty6/Ko2ED4VX5Lk14/QOHDh1aPFN77LHH2jXXXLPSNcyaNavYEaGWF0+88847ttZaaxXnGj9+fLGFWGPHWWedVbxlrX///sXd4FqOo446yq6//vrisYi//vWvtQxtthbhRXhLg6niJ2LRqniDWzk9+GhlgBUeDhsIr4p3qcJb92phv1PqjzX4yyDqH75frtfU+mph32Fh5syZxVZmLtRdu3ZtcF7fZcFl1YV6xf19Xa5X3Je3/mB/JthF3b8AF3n+t7mgEV6EV/2LmL2ORSs7Ac3PHz7go6kEYAPhVf92lCq8vg3GoEGDbMaMGbbddtsVd039n7537tixY4stwFxAJ0yYYKNGjWpwjXVSO2LECJs4cWKDn/kb2VxG/dh7772LO7n+BTb/dqa/TvjUU0+1Rx55pPj5vffea3vsscfy8f4FtunTp9thhx1WiG3dtmV+5/jmm28uvgz37rvvmu/l63d3a9mWrKWQEV6EtyVG+Pm/E2DRgoTmEoAP+EB4NQbwjqZzKlV4/WP8+Vx/41ndYwX+AooFCxaYby3mh28vdtVVV610Rc0J79KlS4s9fl1Q6w7fqcH/vW8n5ofvO+dbivmjDfWPH/zgB/bDH/5w+b/y6/E7xC65dS+e+MxnPmO//vWvCzkv8wA8hLdMnqp8LoSmyt1t/dzgo/UZVvUMsNGws3jHKhRe/yiXXd9KzEF87bXXzB9x8H1zTzzxRBs+fHijV9Oc8NYNuPPOO4u7v/4aY9+CzB9V+NSnPmW77bZbce66VxvX/wAX8P/5n/+xhx9+2P7+97+bfwnO70T7c8HbbrutHXDAAeZ3lf0ayz4AD+Etm6mqno9Fq6qdLWde8FFOjlU8C2wgvCrXpd/hVT84Qx3Ci/Bm4LyMObJolZFidc8BH9XtbWtnBhsIr8oQwqsmFahDeBHeADYph7BopWy7PGn4kKNKVwgbCK8KPcKrJhWoQ3gR3gA2KYewaKVsuzxp+JCjSlcIGwivCj3CqyYVqEN4Ed4ANimHsGilbLs8afiQo0pXCBsIrwo9wqsmFahDeBHeADYph7BopWy7PGn4kKNKVwgbCK8KPcKrJhWoQ3gR3gA2KYewaKVsuzxp+JCjSlcIGwivCj3CqyYVqEN4Ed4ANimHsGilbLs8afiQo0pXCBsIrwo9wqsmFahDeBHeADYph7BopWy7PGn4kKNKVwgbCK8KPcKrJhWoQ3gR3gA2KYewaKVsuzxp+JCjSlcIGwivCj3CqyYVqEN4Ed4ANimHsGilbLs8afiQo0pXCBsIrwo9wqsmFahDeBHeADYph7BopWy7PGn4kKNKVwgbCK8KPcKrJhWoQ3gR3gA2KYewaKVsuzxp+JCjSlcIGwivCj3CqyYVqEN4Ed4ANimHsGilbLs8afiQo0pXCBsIrwo9wqsmFahDeBHeADYph7BopWy7PGn4cx/J3gAAIABJREFUkKNKVwgbCK8KPcKrJhWoQ3gR3gA2KYewaKVsuzxp+JCjSlcIGwivCj3CqyYVqEN4Ed4ANimHsGilbLs8afiQo0pXCBsIrwo9wqsmFahDeBHeADYph7BopWy7PGn4kKNKVwgbCK8KPcKrJhWoQ3gR3gA2KYewaKVsuzxp+JCjSlcIGwivCj3CqyYVqEN4Ed4ANimHsGilbLs8afiQo0pXCBsIrwo9wqsmFahDeBHeADYph7BopWy7PGn4kKNKVwgbCK8KPcKrJhWoQ3gR3gA2KYewaKVsuzxp+JCjSlcIGwivCj3CqyYVqEN4Ed4ANimHsGilbLs8afiQo0pXCBsIrwo9wqsmFahDeBHeADYph7BopWy7PGn4kKNKVwgbCK8KPcKrJhWoQ3gR3gA2KYewaKVsuzxp+JCjSlcIGwivCj3CqyYVqEN4Ed4ANimHsGilbLs8afiQo0pXCBsIrwo9wqsmFahDeBHeADYph7BopWy7PGn4kKNKVwgbCK8KPcKrJhWoQ3gR3gA2KYewaKVsuzxp+JCjSlcIGwivCj3CqyYVqEN4Ed4ANimHsGilbLs8afiQo0pXCBsIrwo9wqsmFahDeBHeADYph7BopWy7PGn4kKNKVwgbCK8KPcKrJhWoQ3gR3gA2KYewaKVsuzxp+JCjSlcIGwivCj3CqyYVqEN4Ed4ANimHsGilbLs8afiQo0pXCBsIrwo9wqsmFahDeBHeADYph7BopWy7PGn4kKNKVwgbCK8KPcKrJhWoQ3gR3gA2KYewaKVsuzxp+JCjSlcIGwivCj3CqyYVqEN4Ed4ANimHsGilbLs8afiQo0pXCBsIrwo9wqsmFahDeBHeADYph7BopWy7PGn4kKNKVwgbCK8KPcKrJhWoQ3gR3gA2KYewaKVsuzxp+JCjSlcIGwivCj3CqyYVqEN4Ed4ANimHsGilbLs8afiQo0pXCBsIrwo9wqsmFahDeBHeADYph7BopWy7PGn4kKNKVwgbCK8KPcKrJhWoQ3gR3gA2KYewaKVsuzxp+JCjSlcIGwivCj3CqyYVqEN4Ed4ANimHsGilbLs8afiQo0pXCBsIrwo9wqsmFahDeBHeADYph7BopWy7PGn4kKNKVwgbCK8KPcKrJhWoQ3gR3gA2KYewaKVsuzxp+JCjSlcIGwivCj3CqyYVqEN4Ed4ANimHsGilbLs8afiQo0pXCBsIrwo9wqsmFahDeBHeADYph7BopWy7PGn4kKNKVwgbCK8KPcKrJhWoQ3gR3gA2KYewaKVsuzxp+JCjSlcIGwivCj3CqyYVqEN4Ed4ANimHsGilbLs8afiQo0pXCBsIrwo9wqsmFahDeBHeADYph7BopWy7PGn4kKNKVwgbCK8KPcKrJhWoQ3gR3gA2KYewaKVsuzxp+JCjSlcIGwivCj3CqyYVqEN4Ed4ANimHsGilbLs8afiQo0pXCBsIrwo9wqsmFahDeBHeADYph7BopWy7PGn4kKNKVwgbCK8KPcKrJhWoQ3gR3gA2KYewaKVsuzxp+JCjSlcIGwivCj3CqyYVqEN4Ed4ANimHsGilbLs8afiQo0pXCBsIrwo9wqsmFahDeBHeADYph7BopWy7PGn4kKNKVwgbCK8KPcKrJhWoQ3gR3gA2KYewaKVsuzxp+JCjSlcIGwivCn2bCO+cOXNszJgxdtddd9msWbOsV69eNmTIEDvhhBNs+PDh6rWtVPenP/3Jxo8fbw8//LC9+eab1rVrV9too43sC1/4gp100kk2ePDgZs/9wAMP2JVXXmmPPvqovfvuu7b++uvbV7/6VTv77LNts802C19XUwMRXoS3dKgqekIWrYo2tqRpwUdJQVbwNLCB8KpYly6806ZNs2HDhtns2bOLa+jdu7ctWLDAli5dWvz55JNPtrFjx6rXt7zuvPPOs4suumj5n3v27Fmcc+HChcW/c/m9+uqr7dvf/naj57700ksLsV22bJl16dKluK733ntv+TX6X5ovfvGLNV9XcwMQXoS3VKAqfDIWrQo3t4SpwUcJIVb0FLCB8Kpolyq8Lp+DBg2yGTNm2DbbbGOTJ08u7rq68F5xxRV2/vnnF8I5YcIEGzVqlHqNdu+999qee+5Z1B9yyCH24x//2D7zmc8U53rmmWfslFNOMb/7261bN3PhHjBgQINz//73v7evfe1rRb3fCf6v//ov69evn/3973+3Y4891vzO71prrWXTp0+3tddeW76ulgoRXoS3JUb4+b8TYNGChOYSgA/4aCoB2EB41b8dpQrvuHHjiju4fvf1hRdesE022aTBdYwePbq4C7vhhhvaK6+8Yt27d5euc+TIkTZp0iTbfPPNi/O62NY/5s6dWzzaMH/+fPM7uWeccUaDn++444725JNP2j777LN8Ya0r8Lu8W221lf3jH/8oxvn4sg6EF+Eti6Wqn4dFq+odbt384KN1+VV5NGwgvCrfpQrvLrvsYlOnTrVjjjnGrr322pWu4dVXX7VNN920uNN6zz33LL9r29LF+t1Zrz/ooIPstttua7S8Tmp/8IMf2Pe///3lNS+++GIhtH78+c9/tt12222l8T/60Y+Ku8+f+tSn7LXXXiseeSjjQHgR3jI4ynAOFq0MXY7PET7i2VV9JGwgvCrjpQmv313t06dPIbO33nqrHXzwwY1ew9Zbb23PP/+8nXnmmXbJJZdI1/md73zHfv7zn9sWW2xRjG3sDu/GG29s8+bNK74ot++++y4/r99R9jvLfm1vv/128azvisdjjz1mO++8c/Gv/fx1gixdXDNFCC/C21qGsoxn0crS6dg84SOWW4ZRsIHwqpyXJrz1pdEfOxg4cGCj13DooYfaLbfc0ujjBU1d9FNPPWV+B9dl2sdffPHFy5/hffbZZ4vHKPwZXv+y3B/+8IcGd2hPPPFE+9nPfmZ+99l3Z2js+OCDD4rHMPzwO8h+J7mMA+FFeMvgKMM5WLQydDk+R/iIZ1f1kbCB8KqMlya8fmd1//33Lz7Xn6n1O6qNHaeddlqxNZgL7OOPP65eZ/EFuOOOO85cTv2ov0uDby929NFH2wUXXGB1kll34gMPPNB+/etf2wEHHGC33357k5+35pprFluV+XPIfkdYOVb8rBXHLFq0qLgb3dznKp9ThZqPPvqomMZqq61Whekwh5ITgI+SA63Y6eCjYg0tcTqw0TBMdx5fZ+t2sCox6k5/qtKE98Ybb7TDDz+8CGTx4sUrPXZQl9S5555b3KH1nRReeumlmgK877777IgjjrA33nijwTjfceHII4+0Cy+80Pr27dvgZ767g+/y4Nfm0tzU4c/v+hfX/Nq+973vSdeF8EoxFUX8UtKzylgJHxm7rs8ZPvSsslXCBsKrMt8phHfJkiXFdmL+HK8/a+tS6i+y8DuoU6ZMKZ4Hdnn2rdAeeuihBtLblsLbUsg80vBxQvxnp5Zoyf1z+Mjd/5ZmDx8tJZT357DRsPd4R9N/F0oT3rZ8pMHf2uZ3Xf25YH+et0ePHg1m5G928y/D+T/95RK+T2/d0ZaPNLT0KwbwEN6WGOHn/06ARQsSmksAPuCjqQRgA+FV/3aUJrz+PO5OO+1UfG7ZX1pbb731ije3Nfd87emnn26XX365bbnlluZbkdUddXv/8qU1FYm2qeOXUtvkWpWzwkdVOtk284CPtsm1CmeFDYRX5bg04a2/LVlzOx34Ywf+NjR1W7J//etfts466xTzufvuu22vvfZqdG5124/5XdUPP/xweY3v0OA7NfiX6N55551GvzRVX9bZlkxFp7Y6finVlle2avjI1vHa5gsfteWVqRo2EF6V99KE1z9w6NChxTO1/rrea665ZqVrmDVrVvH2tVpePOGS6q/99WP8+PF2/PHHNzq3s846q3hLWv/+/Yu7wXVH/RdP+PO9u+6660rjL7roIjvvvPN48YRKTaCOX0qB0BINgY9EzQ5MFT4CoSUZAhsIr4p6qcJb92rhXr16FY81+Msg6h++X67X1PpqYX8728yZM4utzFyoV3x5hL8e2J/hdaFu7PXBdW9h22+//ezOO+9scE3+sgp/0cTrr7/Oq4VVagJ1/FIKhJZoCHwkanZgqvARCC3JENhAeFXUSxVe3/dt0KBBNmPGDNtuu+3s+uuvL/7pe+eOHTvWzjnnnOLu7oQJE2zUqFENrrFOakeMGGETJ05s8DN/I5t/Gc2Pvffeu7iT619g8+1InnzySTv11FPtkUceKX7uW5DtscceDcb//ve/t69+9avFvzvllFPshz/8YbGTg1+n342+//77i7vIvtND3eMTaoDN1fGltY/T4ZdSGURV9xzwUd3eljEz+CgjxWqeAzYQXpXsUoXXP9Sfz/U3ntU9VuDPzi5YsMB8azE/fHuxq666aqXra054ly5dWuyje/PNNy8f5zs1+L/3PX/98I2Wfbsyf7ShscMl2aXZhdtre/fuXbwgww//3/6X5otf/KKam1SH8CK8EigUsUsDDDSbAFIDIE0lABsIr/q3o3Th9Q922fWtxBzE1157zfwRB9831788Nnz48EavrTnhrRvgjyP43V9/jbFvQdalS5fiudvddtutOLfv0dvc8cADDxRvefO7wf5WtQ022MB8n17f8myzzTZTM5PrEF6EV4YleSGLVnIAWpg+fMAHwqsxgHc0nVObCK/WlupXAR7CW33Ky5khQlNOjlU9C3xUtbOtnxdscIdXpQjhVZMK1CG8CG8Am5RDWLRStl2eNHzIUaUrhA2EV4Ue4VWTCtQhvAhvAJuUQ1i0UrZdnjR8yFGlK4QNhFeFHuFVkwrUIbwIbwCblENYtFK2XZ40fMhRpSuEDYRXhR7hVZMK1CG8CG8Am5RDWLRStl2eNHzIUaUrhA2EV4Ue4VWTCtQhvAhvAJuUQ1i0UrZdnjR8yFGlK4QNhFeFHuFVkwrUIbwIbwCblENYtFK2XZ40fMhRpSuEDYRXhR7hVZMK1CG8CG8Am5RDWLRStl2eNHzIUaUrhA2EV4Ue4VWTCtQhvAhvAJuUQ1i0UrZdnjR8yFGlK4QNhFeFHuFVkwrUIbwIbwCblENYtFK2XZ40fMhRpSuEDYRXhR7hVZMK1CG8CG8Am5RDWLRStl2eNHzIUaUrhA2EV4Ue4VWTCtQhvAhvAJuUQ1i0UrZdnjR8yFGlK4QNhFeFHuFVkwrUIbwIbwCblENYtFK2XZ40fMhRpSuEDYRXhR7hVZMK1CG8CG8Am5RDWLRStl2eNHzIUaUrhA2EV4Ue4VWTCtQhvAhvAJuUQ1i0UrZdnjR8yFGlK4QNhFeFHuFVkwrUIbwIbwCblENYtFK2XZ40fMhRpSuEDYRXhR7hVZMK1CG8CG8Am5RDWLRStl2eNHzIUaUrhA2EV4Ue4VWTCtQhvAhvAJuUQ1i0UrZdnjR8yFGlK4QNhFeFHuFVkwrUIbwIbwCblENYtFK2XZ40fMhRpSuEDYRXhR7hVZMK1CG8CG8Am5RDWLRStl2eNHzIUaUrhA2EV4Ue4VWTCtQhvAhvAJuUQ1i0UrZdnjR8yFGlK4QNhFeFHuFVkwrUIbwIbwCblENYtFK2XZ40fMhRpSuEDYRXhR7hVZMK1CG8CG8Am5RDWLRStl2eNHzIUaUrhA2EV4Ue4VWTCtQhvAhvAJuUQ1i0UrZdnjR8yFGlK4QNhFeFHuFVkwrUIbwIbwCblENYtFK2XZ40fMhRpSuEDYRXhR7hVZMK1CG8CG8Am5RDWLRStl2eNHzIUaUrhA2EV4Ue4VWTCtQhvAhvAJuUQ1i0UrZdnjR8yFGlK4QNhFeFHuFVkwrUIbwIbwCblENYtFK2XZ40fMhRpSuEDYRXhR7hVZMK1CG8CG8Am5RDWLRStl2eNHzIUaUrhA2EV4Ue4VWTCtQhvAhvAJuUQ1i0UrZdnjR8yFGlK4QNhFeFHuFVkwrUIbwIbwCblENYtFK2XZ40fMhRpSuEDYRXhR7hVZMK1CG8CG8Am5RDWLRStl2eNHzIUaUrhA2EV4Ue4VWTCtQhvAhvAJuUQ1i0UrZdnjR8yFGlK4QNhFeFHuFVkwrUIbwIbwCblENYtFK2XZ40fMhRpSuEDYRXhR7hVZMK1CG8CG8Am5RDWLRStl2eNHzIUaUrhA2EV4Ue4VWTCtQhvAhvAJuUQ1i0UrZdnjR8yFGlK4QNhFeFHuFVkwrUIbwIbwCblENYtFK2XZ40fMhRpSuEDYRXhR7hVZMK1CG8CG8Am5RDWLRStl2eNHzIUaUrhA2EV4Ue4VWTCtQhvAhvAJuUQ1i0UrZdnjR8yFGlK4QNhFeFHuFVkwrUIbwIbwCblENYtFK2XZ40fMhRpSuEDYRXhR7hVZMK1CG8CG8Am5RDWLRStl2eNHzIUaUrhA2EV4Ue4VWTCtQhvAhvAJuUQ1i0UrZdnjR8yFGlK4QNhFeFHuFVkwrUIbwIbwCblENYtFK2XZ40fMhRpSuEDYRXhR7hVZMK1CG8CG8Am5RDWLRStl2eNHzIUaUrhA2EV4Ue4VWTCtQhvAhvAJuUQ1i0UrZdnjR8yFGlK4QNhFeFHuFVkwrUIbwIbwCblENYtFK2XZ40fMhRpSuEDYRXhR7hVZMK1CG8CG8Am5RDWLRStl2eNHzIUaUrhA2EV4Ue4VWTCtQhvAhvAJuUQ1i0UrZdnjR8yFGlK4QNhFeFHuFVkwrUIbwIbwCblENYtFK2XZ40fMhRpSuEDYRXhR7hVZMK1CG8CG8Am5RDWLRStl2eNHzIUaUrhA2EV4Ue4VWTCtQhvAhvAJuUQ1i0UrZdnjR8yFGlK4QNhFeFHuFVkwrUIbwIbwCblENYtFK2XZ40fMhRpSuEDYRXhR7hVZMK1CG8CG8Am5RDWLRStl2eNHzIUaUrhA2EV4W+TYR3zpw5NmbMGLvrrrts1qxZ1qtXLxsyZIidcMIJNnz4cPXaltftvvvu9uCDD0rjRo4cadddd12DWv93kyZNanb81ltvbX/961+lz1CLEF6EV2Ulex2LVnYCmp8/fMBHUwnABsKr/u0oXXinTZtmw4YNs9mzZxfX0Lt3b1uwYIEtXbq0+PPJJ59sY8eOVa+vqDvwwAPt4YcfbnLMokWL7J133il+Pm7cOBs9enSjwtujRw/r27dvo+cZOHCgPfDAAzVdV0vFCC/C2xIj/PzfCbBoQUJzCcAHfCC8GgN4R9M5lSq8CxcutEGDBtmMGTNsm222scmTJ9vgwYML4b3iiivs/PPPt2XLltmECRNs1KhRWveEqosvvtjOPfdcW3311e2NN96wtdZaq1HhHTFihE2cOFE4YzklgIfwlkNS9c+C0FS/x62ZIXy0Jr1qj4WNhv3FO1aR8PrdVb+D27NnT3vhhRdsk002afDJfuf16quvtg033NBeeeUV6969eyl/E7fcckubPn26HXzwwXbrrbeudM66RxoQ3lLiDp2EX0qh2NIMgo80rQ5NFD5CsaUYBBsIrwp6qXd4d9llF5s6daodc8wxdu211650Da+++qptuummxV3ee+65x/bcc0/1Opus80cddt111+Lnd999t+21114Ib6tTLf8E/FIqP9MqnRE+qtTN8ucCH+VnWpUzwgbCq7JcmvDOnz/f+vTpU8is32X1u62NHf7lsOeff97OPPNMu+SSS9TrbLLuuOOOs2uuucY22GADe+2116xr164Ib6tTLf8E/FIqP9MqnRE+qtTN8ucCH+VnWpUzwgbCq7JcmvA+9thjtvPOOxef648z+JfAGjsOPfRQu+WWW2yfffZZ/kUV9WJXrPvggw8K0Z07d26zAl33SMNGG21k/sU1v9Ps/9x8882LO8L+qMV6660XvYwmx/EszcfR8EupdLwqdUL4qFQ7S58MfJQeaWVOCBsIrwpzacLrW5Dtv//+xee6gPrd3saO0047za688krbcccd7fHHH1evs9G6G2+80Q4//PDiZ37XeKuttmq0rv62ZH4H2K/Nr/Gjjz4q6v1Lbi7hX/7yl1t1PSsORngR3lKBqvDJWLQq3NwSpgYfJYRY0VPABsKrol2a8NaXz8WLF1u3bt0avQbfTcF3VRgwYIC99NJL6nU2WveVr3zF/vCHP5g/O/zoo482ea6rrrrKVlttteJu7qc//enisYd58+YV+wSfccYZxc4Oa6yxhj3xxBPFdalHndA2Ve/bpXkOt99+u3rKytbV/T8X3gcOElgxAfiAieYSgA/4aCoB2GiYjG/j6uus75rF0TCBTiu8/ryufwHOYR8/frwdf/zxod76bhH+Ugzfx/eb3/ymubirB8KrJmXL76YjvHpmmSpZtDJ1u/a5wkftmWUZARsIr8p6acK7qh9puOiii+y8884rnsX1O7T9+vVT57xSne8P/KMf/ah41MHFtywp45GGj6PmPzuF8UwxED5StDk8SfgIR1f5gbDRsMV4R9PIlya8/jzuTjvtVHzSqvjSmj968PLLL9thhx1mN910U6v+Uvt2Zv4lOj/8DXH9+/dv1fnqBgMewlsKSAlOwqKVoMmtmCJ8tCK8ig+FDYRXRbw04a2/Ldltt91mBx10UKPX4G9g89cPt2Zbsr/85S+22267FecvYz/f+sI7Z84cW2edddT8mq1DeBHeUkBKcBIWrQRNbsUU4aMV4VV8KGwgvCripQmvf+DQoUNtypQpduyxxxZ74654zJo1q3j7WmtfPOHn/+Uvf2m+zdjMmTNb/QjCBRdcYBdeeKH17t3b3n333Vafjzu8K+PHLyX1r2TOOvjI2Xd11vChJpWvDjYQXpX6UoW37tXCvXr1Kh5r2HjjjRtch7922Gta82ph33t3/fXXt/fee8/OOecc82d5mztcrrt06dJkie/Ju8MOO9jbb79dyuMR9T+IO7zc4VX/ImavY9HKTkDz84cP+GgqAdhAeNW/HaUKr2+DMWjQIJsxY4Ztt912dv311xf/dEkdO3ZsIaguoBMmTLBRo0Y1uEbfccHv1o4YMcImTpzY5PVPnjzZjjzyyOLn06dPty222KLZufo13HHHHcUYfwyi7nEFfwTD/6L4tmSvv/66uaT7tmRbbrmlml2LdQgvwtsiJBQUCbBoAUJzCcAHfCC8GgN4R9M5lSq8/jH+fO6wYcOKL3/54TsfLFiwwJYsWVL8+aSTTjLfF3fFQxXePfbYw/74xz/arrvuag899FCLBLg815dr32/XgfDdGOq2M/Evqd188832pS99qcXz1VIAeAhvLbxkrkVoMne/5bnDR8sZZa2AjYadxztWofD6R7nsjhkzprhr4/vl+t1T3+v2xBNPtOHDhzd6NYrw1t97158R9md5Wzr8rrHfFfYvuvmLLt56661CwPv27Vvcjd57773tW9/6VvG2tbIPwEN4y2aqqudj0apqZ8uZF3yUk2MVzwIbCK/Kdel3eNUPzlCH8CK8GTgvY44sWmWkWN1zwEd1e9vamcEGwqsyhPCqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWimYP4fwAAAgAElEQVTbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0KP8KpJBeoQXoQ3gE3KISxaKdsuTxo+5KjSFcIGwqtCj/CqSQXqEF6EN4BNyiEsWinbLk8aPuSo0hXCBsKrQo/wqkkF6hBehDeATcohLFop2y5PGj7kqNIVwgbCq0LfJsI7Z84cGzNmjN111102a9Ys69Wrlw0ZMsROOOEEGz58uHpty+t23313e/DBB6VxI0eOtOuuu67R2qefftp+8pOf2AMPPGBvvfWW9e/f3/zcZ5xxhg0ePFg6fy1FCC/CWwsvmWtZtDJ3v+W5w0fLGWWtgA2EV2W/dOGdNm2aDRs2zGbPnl1cQ+/evW3BggW2dOnS4s8nn3yyjR07Vr2+ou7AAw+0hx9+uMkxixYtsnfeeaf4+bhx42z06NEr1d54443mMrx48eLiZ3379rW5c+cW/3v11Ve3SZMm2WGHHVbTdbVUjPAivC0xws//nQCLFiQ0lwB8wEdTCcAGwqv+7ShVeBcuXGiDBg2yGTNm2DbbbGOTJ08u7py68F5xxRV2/vnn27Jly2zChAk2atQo9RpbrLv44ovt3HPPLcT1jTfesLXWWqvBGJdwv8PsYnzIIYfYlVdeaRtssEFRe8opp9itt95qLqd+B3jgwIEtfp5agPAivCor2etYtLIT0Pz84QM+EF6NAbyj6ZxKFV6/u+p3cHv27GkvvPCCbbLJJg0+2e+8Xn311bbhhhvaK6+8Yt27d9c62ELVlltuadOnT7eDDz64kNcVj4MOOshuv/1223777e2xxx6zbt26LS9ZsmSJ7bTTToXsfuMb37BbbrmllGvykwAewlsaTBU/EUJT8Qa3cnrw0coAKzwcNrjDq+JdqvDusssuNnXqVDvmmGPs2muvXekaXn31Vdt0002Lu7z33HOP7bnnnup1Nlnnjzrsuuuuxc/vvvtu22uvvRrU+mML6667bnF39/rrr7cjjjhipXP5negjjzyyEFR//tgfwyjjQHgR3jI4ynAOFq0MXY7PET7i2VV9JGwgvCrjpQnv/PnzrU+fPoXM+l1Wv9va2LH11lvb888/b2eeeaZdcskl6nU2WXfcccfZNddcUzyi8Nprr1nXrl0b1LoE77PPPsW/8+eK/YtqKx4uuS7Ffvzud7+zr3/9662+Lj8BwovwlgJSgpOwaCVociumCB+tCK/iQ2ED4VURL014/VGBnXfeufhcf5yhqWdhDz300OKxAZfQOlDVi12x7oMPPihE1+/iNiXQviuD/2y99dazN998s8mP8p+7EP/0pz+1008/PXpJDcYhvAhvKSAlOAmLVoImt2KK8NGK8Co+FDYQXhXx0oTXtyDbf//9i891AfW7vY0dp512WvGlsR133NEef/xx9TobrfOdFw4//PDiZ37XeKuttlqp7rvf/W7xhbkddtjBnnzyySY/z3/uz/G67Lr0lnEgvAhvGRxlOAeLVoYux+cIH/Hsqj4SNhBelfHShLe+fPrWX/W/GFb/Ynw3Bd9VYcCAAfbSSy+p19lo3Ve+8hX7wx/+YP7s8KOPPtpoTd0jD/6c70MPPdTk5/nP/Xlgr//FL34hXVed0DZV7M8New7+hbnsx0cffVREsNpqq2WPgvk3kgB8gEVzCcAHfDSVAGw0TMa3cfV11nfN4miYQKcVXn9e178A57CPHz/ejj/+eIS3A9PNL6UO3JwOcGnw0QGa0IEvAT46cHPa+dJgA+FVESxNeFf1Iw0XXXSRnXfeedajR49iP91+/fo1OmceaVBRaNs6/rNT2+bb2c8OH529g217/fDRtvl25rPDRsPu8Shl0zSXJrz+PK7vZ+vHqvjSmj8S8fLLLxdvR7vpppuanKE/j+uvDuZLa+37K41fSu2bf0f/dPjo6B1q3+uDj/bNvyN/OmwgvCqfpQlv/W3JbrvtNvOXPTR2+BvY/M1nrdmW7C9/+Yvttttuxelb2s/Xtxnbe++9i1rffmydddZZ6bLYlkzFJV7HL6V4dhlGwkeGLsfnCB/x7Ko+EjYQXpXx0oTXP3Do0KE2ZcoUO/bYY4u9cVc8Zs2aVbx9rbUvnvDz//KXv7SNNtrIZs6c2ewXod57771i713/Apm/YKJuV4f613bDDTcUL6TgxRMqNrXX8Uup9swyjYCPTN2ufa7wUXtmWUbABsKrsl6q8Na9WrhXr17FYw0bb7xxg+vw1w57TWteLex7766//vrmInvOOeeYP8vb0lH3auEhQ4YUb4Kr/3IKf7Ww7/LgW5bxauGWkoz/nF9K8ewyjISPDF2OzxE+4tlVfSRsILwq46UKr2+DMWjQIJsxY4Ztt912xat8/Z8uqWPHji0E1e/uTpgwwUaNGtXgGn3HBb9bO2LECJs4cWKT11/3GmAvmD59um2xxRYtztUfoXDZ9bu8/uIL3wfYpdlfRHHqqafazTffXNzd9X14m3phRosf0kgBD49/HAq/lCIE5RkDH3l6HZkpfERSyzEGNhBelfRShdc/1OVy2LBhxVvL/PAXUCxYsMD8TqofJ510kl111VUrXZ8qvHvssYf98Y9/tJb21V3xA3yf4JEjR5rvEdylSxfr27dv8YIMF/Du3bvbr371q+ILcGUeCC/CWyZPVT4Xi1aVu9v6ucFH6zOs6hlgA+FV2S5deP2DXXbHjBlTvDrY98v1Rxz8DuuJJ55ow4cPb/TaFOGtv/euPyPsz/LWcvgd3EsvvdQefPBBe+utt4ovsO2+++7FF+gGDx5cy6mkWoQX4ZVAoWj5a8b33Xdf0iCBlRJAaoCiqQRgA+FV/3a0ifCqH171OoQX4a0642XNj0WrrCSreR74qGZfy5gVbCC8KkcIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzCopWy7fKk4UOOKl0hbCC8KvQIr5pUoA7hRXgD2KQcwqKVsu3ypOFDjipdIWwgvCr0CK+aVKAO4UV4A9ikHMKilbLt8qThQ44qXSFsILwq9AivmlSgDuFFeAPYpBzColXdts+fP9+mTZsWnuAHi5bYb+5/pBi/75c+a59cvVvoXFtvvbWtscYaobEM6rgJ8LsD4VXpRHjVpAJ1CC/CG8Am5RAWreq2fcqUKTZ06NB2n+Cjjz5qu+yyS7tfBxdQbgL87kB4VaIQXjWpQB3Ci/AGsEk5hEWrum1HeKvb244wM353ILwqhwivmlSgDuFFeAPYpBzCotXx2/7eh4vt77Pn13yhC96fb3+f/mLN4+oGvPr2Arv6/r8XfzzxS/9hm6zVM3Su/xgw0Hr2qv2Rhv9Ydw3r06N76DMZ1PYJ8LsD4VUpQ3jVpAJ1CC/CG8Am5RAWrY7ddpfd3cbcZ+99uKRjX2gbXF2fHt3sobOHIb1tkG0Zp+R3B8KrcoTwqkkF6hBehDeATcohLFodu+1PvfqOHfCzhzv2Rbbh1f36hM/ZDpus2YafwKmjCfC7A+FV2UF41aQCdQgvwhvAJuUQFq2O3fb6wnvJQdvagPV6r7IL/mDxUvvt/9ulYR/fpaF711Xy2dP/Oc/O+t/PFZ+F8K6SyEMfwu8OhFcFB+FVkwrUIbwIbwCblENYtDp22x+Z8bod/qu7iou89KDtbMD6q054/TMfeuih4rN32223VRbU9Dfn2Zn/+9ni8244aj/77Gc+tco+mw/SE+B3B8Kr0oLwqkkF6hBehDeATcohLFodu+23PfcX++GTx3fsi2zDq/v+kJ/bwdvu2oafwKmjCfC7A+FV2UF41aQCdQgvwhvAJuUQFq2O3XaEF+HtqITyuwPhVdlEeNWkAnUIL8IbwCblEBatjt12HmngkYaOSii/OxBelU2EV00qUIfwIrwBbFIOYdHq2G2v/6W19vgCV3vw0d5z7thEdJyraw82Os7sV74SvKPp7iC8bUgu4CG8bYhXpU7NotWx29ne8tcefLT3nDs2ER3n6tqDjY4ze4S3ll4gvLWkVWMtwovw1ohM2nIWrY7d+vaWv/bgo73n3LGJ6DhX1x5sdJzZI7y19ALhrSWtGmsRXoS3RmTSlrNodezWt7f8tQcf7T3njk1Ex7m69mCj48we4a2lFwhvLWnVWIvwIrw1IpO2nEWrY7e+veWvPfho7zl3bCI6ztW1BxsdZ/YIby29QHhrSavGWoQX4a0RmbTlLFodu/XtLX/twUd7z7ljE9Fxrq492Og4s0d4a+kFwltLWjXWIrwIb43IpC1n0erYrf/L396yw6+dUlzk9/cdZNtv3G+VXTCvFl5lUXfKD+J3R8O24R1NY4zwtuFfccBDeNsQr0qdmkWrY7fzur/8X/vhb57v2BfZhlfXHluxteF0KnVqfncgvCrQCK+aVKAO4UV4A9ikHMKi1bHbnll4+/ToZg+dPcz69OjesZuU9Or43YHwqugjvGpSgTqEF+ENYJNyCItWx2777Hkf2t3PvlFc5ID1elvP1buusgue/s959v+3dybgdk3nH/4QISShhghFzXNoo2Y1xFxUCEUVTWlp0JpnlWhJgjbGR2tItdQclLRFB0WLoNI2pkSlhphrShEi+D/v6n/d7ntyzj3r7LPOuXfv9VvPkye5uWvvvdb7fXut3/72t9Y+YcIUd72xwwa567ezrDygr8RuO4E3eC2NHRK8oS4jwRtKKkc9CV4J3hxuk+QhmrSSNHtQp7V4LAhTspU0dkjwhjq/BG8oqRz1JHgleHO4TZKHaNJK0uxBnZbgDcKUbCWNHRK8oc4vwRtKKkc9CV4J3hxuk+QhmrSSNHtQpyV4gzAlW0ljhwRvqPNL8IaSylFPgleCN4fbJHmIJq0kzR7UaQneIEzJVtLYIcEb6vwSvKGkctST4JXgzeE2SR6iSStJswd1WoI3CFOylTR2SPCGOr8EbyipHPUkeCV4c7hNkodo0krS7EGdluANwpRsJY0dEryhzi/BG0oqRz0JXgneHG6T5CGatJI0e1CnJXiDMCVbSWOHBG+o80vwhpLKUU+CV4I3h9skeYgmrfKa/d1337XHH388dwenvjKz0z68qw/sn+tca6+9tvXt2zfXsTqo5xLQ2CHBG+qdEryhpHLUk+CV4M3hNkkeokmrvGafNGmSbbzxxt3ewQcffNA22mijbm+HGhCXgMYOCd5Qj5LgDSWVo54ErwRvDrdJ8hBNWuU1uwRveW3bE3qmsUOCN9QPJXhDSeWoJ8ErwZvDbZI8RJNWec3ebErDrNlz7Pa7H3CAdt16E+vTu1cuWEppyIWtxx+ksUOCN9RJJXhDSeWoJ8ErwZvDbZI8RJNWkmYP7rT8IxhVchXlGxK8oU7fEsH7+uuv25gxY+y2226zGTNm2MILL2yDBw+2ESNG2NChQ0PbVrXee++9Z5dccondfPPNNm3aNCN6sNRSS9laa61l2267rR1zzDFzHbfVVlvZPffc0+V1d955Z5s4cWJTbas8WIJXgjeqQ5X4ZJq0SmzcCF2Tf0SAWNJTyDckeENdO7rgZTXukCFD7LXXXnNt6Nevn73//vv28ccfu5+/+93v2vnnnx/avk71Jk+e7ATz888/7/5//vnnt4UWWsjeeecd9/N8881nc+bMqSl4Ed61Vulut912dtVVV+VqV62DJHgleKM6VIlPpkmrxMaN0DX5RwSIJT2FfEOCN9S1owreDz/80EVap0+fbuuss45dffXVtt566znBO27cODvttNPs008/tfHjx9vw4cND2+jqPf30026l75tvvmmbbbaZjR492v0977zzGlHfRx991EV9uU5l8RHe008/3UaOHNnQdZupLMErwduM/6R0rCatlKzdeF/lH40zS+UI+YYEb6ivRxW8F154oYvgEnV98sknbfnll+/UjsMPP9wuvvhiW2aZZezZZ591EdrQssUWW9h9991niNc777zTevfuHXqoO4aUBgneYGTRK2pQio60VCeUf5TKnNE7I/+IjrQ0J5RvSPCGOnNUwcsehw899JAddNBBdvnll8/VBlIRVlhhBRflRbRuv/32Qe289957bcstt7R55pnHnnrqKVtttdWCjvOVJHgbwtWSyhqUWoK1NCeVf5TGlC3piPyjJVhLcVL5hgRvqCNHE7wsHuvfv78TszfeeKPtueeeVdvA1jBPPPGEHX/88TZ27Nigdh5yyCF26aWX2uabb+6ivI0WCd5GicWvr0EpPtMynVH+USZrxu+L/CM+07KcUb4hwRvqy9EE78MPP2wbbrihuy7pDGussUbVNuy99952ww032C677GLeUes1dtCgQfbYY4/ZSSedZIceeqidccYZdscddxi7QSyxxBJOCB999NE1v6LjBe8qq6xis2fPtpdfftktXltzzTVtt912c+dErMcuyuH9H1ENSrG9q1znk3+Uy56xeyP/iE20POeTb0jwhnpzNMHLFmSIRwq7JtQSkEcddZSdd955tv7669sjjzwS1M4FF1zQWBB3xBFHuJ0U3n77bZfDS64w/6aweI3dH8gTrizZbcn8cbSRaDSFXGPazwK7RooXtLWOQVz36tXLLaZLvXzyyScddkqdhfo/NwH5h7yiKwLyD/lHLQLyjc5k9thjD6eH0EwqnQlEE7zXXHON7bfffu7sH330kRN61copp5xiZ511lsvDnTp1al17fPDBB9anTx9Xjxxe/s0+vPvss48Tvc8884wddthhLicYI5Pvy+4N2cJiNXKH2XqMBXPUY7eHa6+91k4++WSbOXOmLb300jZlyhRbfPHF67bJV5DgDUZlGpTCWaVYU/6RotXD+yz/CGeVWk35hgRvqM/3eMHLlmbsn+sLUVx2gsiWWbNm2aqrrmovvviiNfoBCaLMm266qRPppEwgxmMVpTT8j6ReO8XyqnKeR/5RTrvG6pX8IxbJ8p1HvtHZptIdtX08muBtZUoDH5TgKY7UBlIRqm1JhlAlekyaAxFbjgkt+++/v9szmL2DifLGKnI8Cd5YvlT282jSKruFm+uf/KM5fmU+Wr4hwRvq39EEL5HSDTbYwF039qK1ZZdd1kVv+agFX3KrVsiTHTZsmPvVq6++agMGDAhl4PYGJvcXscxHLGIVCV4J3li+VPbzaNIqu4Wb65/8ozl+ZT5aviHBG+rf0QRvdluym266qUN8VjaEKCqitZFtyXbccUeXo9uV4J0wYULHVmh81njJJZcMZdAheEmdoB+xigSvBG8sXyr7eTRpld3CzfVP/tEcvzIfLd+Q4A3172iClwvy6d9JkybZwQcfbJdddtlcbZgxY4bbEaHRD0+ce+65dtxxx3WZ0nDmmWfaqaeeav369XM7N7AwLbQccMABbvcH9ghm+7NYRYJXgjeWL5X9PJq0ym7h5von/2iOX5mPlm9I8Ib6d1TB6z8tTKSUtIbllluuUztYbEadRj8t/MILL5jfQ3fcuHF25JFHdjovC9tYtPbSSy+5KC8fvvAFcc3uDrXK5MmTnVBnC7ETTzzRRo8eHcqubj0JXgneuk6iCo6AJi05QlcE5B/yj1oE5BsSvKF3R1TBy75vpB1Mnz7d1l13XRc15W92UWB3BbYAQ4COHz/ehg8f3qmNbBv23HPP2YEHHmhXXnnlXO0nBeKcc87p2JZs33337diWbMSIEXbXXXe5n/m0cXY/XQTstGnT3DZmCNtFFlnEnfutt96y66+/3u3MQER44MCBLrrbyLZk9SBL8Erw1vMR/f6/BDRpyRMkeOUDeQho7JDgDfWbqIKXi5KfO2TIECOPlsIHKIjAzpkzx/3MxyMuuOCCudpXT/By/F577WW33nqrOxYxySIzhKv/GaGMsM2WkSNH2qhRozr+i/awgwMi1394YqWVVrJbbrnFifOYRYJXgjemP5X5XJq0ymzd5vsm/2ieYVnPIN+Q4A317eiClwsjdseMGeOiNqQjkOIwePBg94GIoUOHVm1bPcHrDyJqfMUVV7jtw1hgxgcjttlmGzv22GPdp4IrCwL8uuuus/vvv999pOKNN95wXyBZbLHFjE8W77777i6qnN3rNxRevXoSvBK89XxEv/8vAU1a8oSuCMg/5B+1CMg3JHhD746WCN7Qi5e9ngSvBG/ZfTxW/zRpxSJZzvPIP8pp1xi9km9I8Ib6kQRvKKkc9SR4JXhzuE2Sh2jSStLswZ2WfwSjSq6ifEOCN9TpJXhDSeWo53eHqPZluBynK/Qh+t55oc3X8sbLP1qOuNAXkH8U2nwtbbx8ozNedpyi+DVKLYVfsJNL8LbQYBK8/4Prb0KJ/xY6XIFPLf8osPHa0HT5RxsgF/QS8g0J3lDXleANJaV6TRFQekdT+Ep/sPyj9CZuqoPyj6bwlfpg+UapzRu1cxK8UXHqZLUIaFCSb3RFQP4h/5B/yAfyENDYkYdamsdI8KZp97b3WoNS25EX6oLyj0KZq+2NlX+0HXlhLijfKIypur2hErzdboI0GqBBKQ075+2l/CMvuTSOk3+kYec8vZRv5KGW5jESvGnave291qDUduSFuqD8o1Dmantj5R9tR16YC8o3CmOqbm+oBG+3myCNBmhQSsPOeXsp/8hLLo3j5B9p2DlPL+UbeaileYwEb5p2b3uvNSi1HXmhLij/KJS52t5Y+UfbkRfmgvKNwpiq2xsqwdvtJkijARqU0rBz3l7KP/KSS+M4+Ucads7TS/lGHmppHiPBm6bd1WsREAEREAEREAERSIaABG8yplZHRUAEREAEREAERCBNAhK8adpdvRYBERABERABERCBZAhI8CZjanVUBERABERABERABNIkIMGbpt3VaxEQAREQAREQARFIhoAEbzKmVkdFQAREQAREQAREIE0CErxp2l29FgEREAEREAEREIFkCEjwJmNqdVQEREAEREAEREAE0iQgwZum3Zvq9ZVXXmnDhw+3Lbfc0v70pz81dS4dXC4C8o1y2TNWb2L6xTe+8Q37+c9/bqeffrqNHDkyVhN1nsgEYto8ctN6xOmeffZZW3HFFV1bPv30005twq9HjRplBx54oMFRJQ4BCd44HJM6iwaypMzdUGflGw3hSqZyTL+Q4C2G28S0eTF63FgrJXgb4xWjtgRvDIqJnUMDWWIGb6C78o0GYCVUNaZfSPAWw3Fi2rwYPW6slRK8jfGKUVuCNwbFxM6hgSwxgzfQXflGA7ASqhrTLyR4i+E4MW1ejB431koJ3sZ4xagtwRuDYmLn0ECWmMEb6K58owFYCVWN6RcSvMVwnJg2L0aPG2ulBG9jvGLUluCNQbGbzjFjxgxbfvnlXcL7Aw88YBtvvHHNlqy99tr2xBNP2FlnnWUnnXSSffzxx3bffffZr371K/f3iy++aG+88YZ95jOfsfXXX98OPvhg22OPPaqeL+ZA9u6779odd9zh2vG3v/3NteP999+3gQMH2pe+9CU76qijbPDgwV0Sfumll+y8886zO++80/71r3+5vi277LLuuH322cd22223uY6fM2eO/eIXv7Drr7/eJk+ebO+8844tueSStvLKK9vQoUPtgAMOsMUXX7ybLNv8ZVP3ja222sruuece+9nPfmYIpGqlXh3uq5tvvtn5yUMPPWRvvvmmuz9WWGEF22WXXdyCkuWWW655Y7XxDGXwi3qCd/bs2XbppZfaNddc48a8Dz74wJZZZhnbbrvt7Pjjj3f3eLXy73//284991z7zW9+Y9OnT7ePPvrIllhiCfvsZz9rW2+9tbP3Wmut1enQKVOm2I9//GPna4xD8803nw0YMMBWWmkl23777d042t3jSJlsjg3Gjx9vFydBVFUAABIfSURBVF10kbsvp02bZv/5z3/cuM996cvbb79tF1xwgZtX/vnPf9qHH37o7IgPHHvssbbKKqvUvOveeustu/DCC23ixIn29NNPd/jPOuusY1/96lfdn/nnn7/j+Mcee8xdh/nnueees1deecUWWmghW3PNNW2vvfay73znO7bgggvOdT0J3jYOfP9/KQne9jOPekUGYnZKOPzww91NWq0gJL/whS/YPPPM4wZyBgZu0kGDBnVU7927ty2wwAJu8PDlkEMOsZ/85CdznTKm4GXgOuKIIzqusfDCCxtilAGK0qtXLzfA7b///lX7duutt9rXv/51e++999zv6QN9yfajcgXsyy+/bF/5ylfskUceccfMO++81r9/fyd6fd2uhFJUA7bwZCn7Rj0xC/au6uA/TGw8jFG4dxZZZBHnVzxQUYq6S0DR/aIrwctD+0477WQPP/ywsxFjAWJj5syZ7mf+fd111831EPz888/bpptu6h64KQhXxgSEkx8TTjnlFPvhD3/YccfiGzxMI7D92MP5GUd8+d3vfmfbbrttC+/ysFOXxeYEIrDJbbfd5uaGvn37up+zghfb77rrrvbqq686OIhT/MDPEYhRAh08tFaW+++/3wU8Xn/99Y5jqZ+1aaW45qEIv8vOJbTJly9+8Yv2xz/+0fr169fpchK8Yb4bs5YEb0ya3XCuyy+/3L71rW+5qIKPMFQ247jjjnORi80339xFcyk8GTOAIxY32mgjW2qppdykThSLJ+dTTz3VDRA33HCDe0rNlpiC99prr7W//OUvtu+++7roCRE0JhgGAyInCGImkalTp7podrZMmjTJRYGJxGy22WY2duxY22STTZyAZQBC0P70pz91UTpfmJyoy+8WXXRRO+ecc2zvvfd2gxEim+vceOONxiBVLTLcDSbOfcmUfaNZwYvtmVR5gPrBD35g3/zmN12kDl/joZGIDv7z7W9/O7d9uuvAovtFV4J35513dhFahNDFF1/sxhUEz5NPPuls9ec//9n69Onj3uqsvvrqHSbAvjzkEv294oor3LjCOMJ4wVg0YcIEN8YedNBBHcdQF1/48pe/bD/60Y9sjTXWcL/jrRWR5V/+8pe233772YYbbthdpu64bllsjl0JiJx//vkuCIItmfe4FxGmRLM///nPu/Gfue3EE090duEB5plnnrHvf//7LvLPeYjOZ6PC2JK3m4hV5iLmn2222cYJax6YCBzhI4wHvEH0Zffdd3cPWTvssIOLIlN/1qxZLuLL3EuEecSIEc4fs0WCt/23hQRv+5lHvSI3J6//EWtEHLjpsgXxiFBkILjkkkvs0EMPDbo++1wysSAc7r777pYJ3nqNYb9fBDb7EhJRyxZSOBC9/M0rRZ7i6xUi1rxiYhLkaR5hW9aSsm80I3i5j5jAKLxBKPqDT6V/F90vaglexCxClcJDLkIkW3iAX3fddZ1I/drXvuYEqS8IHEQxD+CkQdUrr732mgsSUHhjxBjck0tZbA5jxHv2wSPL3fsGQSDSWqoV7m3ucd4skvbgC3Yn8sveuI8++qgT0c0WxC6pDcxNpMwg0H2R4G2WbuPHS/A2zqzHHUGu7S233OLyThGq2UK6A6+zEHgMzKH5ZOQxLbbYYi66SsSCJ2RfYkZ468GkX/SP3Ku77rqro/pTTz3lBhLKvffe2zHR1TufF8lEdIjklL2k6BvYtBnB6ye+IUOG2B/+8IdSukiR/aKW4P3e977nBAyi9u9//3tVu/kHXoQHr6l9LiZvhh588EEbN26cHXnkkXVtTgSPKOEnn3ziosVEFXt6KYPNSR9gHiOKWlmwCW8ICf7wBnPVVVetapKrrrrKzZVE+JlHKMxxHOvXdtRKoctj4/XWW8/+8Y9/uDeZpM1I8OahGOcYCd44HLv1LLxu23PPPd1refKWsk+RvMa77LLLXE4Tr2izhcUcPC0jKh9//HGXzsAr28pCPhMDTasELwMYqQvku7FIgNdHTCTZwqI78o4r20BeJeKcdIx6hb7x2otBDWa1FuXVO0+Rfp+ibzQreHnNyeITXlMfffTRRTJ3cFuL7Be1BC/RXaK8LHTldXS1QsTNCyFEMeKYQm7uaaed5iJxRA+HDRvm0qMYL2oVcnN5ICLV4bDDDjPSKRC+2eBAsEHaULEMNt9xxx3tt7/9bVVa2Qi/j75Xq0iaCnMG8yQLpCk+MMS/icSGBob8+XkThJD+61//akT/Ed+VhVQ55mlfFOFtg9NXXEKCt/3Mo1+RJ1pucCIWvJJhsQ2FG5tXbdzcLNQgV9UXhDERLHLNfGFwZ9EYuWsUn/TPjfm5z32uo17MCC9PveTA+UUlXITFIkSWEbF+cEKEsFjAF/J1yc9i5Sy5WCGF/vhXj+Twkq9V9pKib2DTZiK8TIQ8DN50001O+JSxFNkvaglecjXJwWe9wjHHHFPVbNjVBwR+//vfuxxNP1aS75vN90e4stMLwQICB5UiinGR32UfxBlDWSuBsCGKSA54TyllsDl5uQjLagVB6ee+UOZ+QSLzJm92iNpnFzzXOw8LWEmPYa2LLzw0cR7/9sAHkpg32WXCFwneenTj/16CNz7TbjkjUQl2MyDfkKdNCgtrWHFaLfLLTUq+GkKShVuIX1IYfOFG9q+NKlelxhK8RFpZ+MEKaa7PggIWeGQj1ERQiKQguBkgJHgbd6/UfEOCN8xHiugX9Kye4O0qMo/o81tEZQWvJ8aDMKKXxb2s9ve7xTCG8ibMC+TsOMkiOaKORBgRv15EsfUVawvYEq2nlKLbHMHI/FOtENThoYWHDr8jQyj3vIKXPGF2M2LOYstP3hqyVWH2jaN/81C5848Eb6h14tWT4I3HslvPxMIyRCNPl0QySbjnaZen3spBglf7RFGJdtTav5ftefxK1FYJXiaVLbbYwuVOsaiu2utDFpbwVF8peP2iukZSGhDYXIP+p5LSgFOm5hv02b9u7mqhJlv1+ZXX2b16WbTCZFTmlIai+kVXgreZlIZqgzfjI4ubeJNE5BjhSqpLtfxRfzyvwxFPJ598sntrxQ432ehft04SBR0LsjbvSvD6cY76LEzkPg4tPJjwVojSSEoDD0BsOTZ69GjnJ9UKKTSk0kjwhlqjdfUkeFvHtq1nJueV3RgQquTskr7AKzhyiVjsxaIvX9jGhe1TKAzq1V67eUFJnVYJXraHYdseFpIhvKsVv0tDpeBlUYLfVqiRRWssGuBaqSxag2lqvkGfERqkJFTunep9jEUq5KUTwauciPzbjzIvWiuqX3QleFlsxnZVXS1ay0bksovWuhqs/T7m1GGtQ+XHJ6ody4dwyCXmYzbkdPaUUsSxIFTwck/zlpKABilvfGQktBARJvDCsWzLGbpobbXVVnPrTqrtkMS1s1FcCd5Qa7SungRv69i2/cx+v12eVIlY8YecVaKn2YUUDAxEeHn1xrZelftEksNE7hpPpa0UvL/+9a/d5t8MUgj1yq/RMNGwRzB5vJWCl3Z58drItmR+wiMSTv5wmbclyzpgar5xxhlnuG3sWOxIjnflosZRo0a5re4olRMRD4h+ez/SgvhISVlL0fyiK8HL/Uz+LKXadnIsUGLFPOMaD9pXX311h1kZY2pta0h01++x63dk6Ko+J2X/b7aA5A1U9iMEPcGPimzzriK8sMWuBFKYU9haLLv2pJI9a1sQub74B91GtiXbYIMN3J7utQS23/Gl2jijlIb23w0SvO1n3rIr+kgEi86IcvAzUQ+22qksTAxMEAgCorl+ARfRT77axqs7//WYVkV4mQhIm+DpmtwnthQi8sxkwoRFO4hI0I5qgpfPvdIP/+GJs88+262s9h/QIKeObYjIsfOl8sMTLHAh9cN/eIJtasgRI9Wich/PlhmuDSdOzTfYUxXf5qGOfZfPPPNMN7mxgAQ/Y1U+NscHq31Vj9x3xC5vP6jLmwZWbpMWwwb2pNrwijt0X+s2mDjXJYrmF10JXn7HAzQP0tiWdBb/GVjua2zFq+tqH55Yeuml3WeAebhBFHvxCx92YGDPbsYmRAopDazqZ2wlJ5adA8jXZdzBP3hg4v/5xCxv2sgt7UmlyDavJ3hfeOEFN5exsxD3J3MC9zKLsSkEVsjdJvDBgsNsGgLzHGlORP4ZO0hp8h+eIAhETjcPMvy/T/fjA02MLfgb8wb+g39wLt4uYXvSCxHXivB2/10gwdv9NojaAm7U7M4L3KTVopj8P5Fgvy2Lz5/lZyYEJnu+BU9pleDl3Hx9BmHrCxER2oCIReQymFTL4fX1WUjCU73fBoYoMZNVdteHap8WZgshojUUHhC4LgOd3w6tDJ8WrnSs1HyDV5osyPSFicdveYdfIUwQQNVszQTHSnu/9zNixh9f9E8LF90v6n1aGAHqPxvOWOD33KXfPMCQY1v5MZHsGwDehjEe8CbMfzaYVfe33357R55ndhsrzutX5vMA5ccQXneTV9qTFq152xdtLPA2ryd46R+CHpFL0MaP79y7pO/5+Y7/HzNmjJ1wwgmdbgeCQPiGD/aw0wJiORulz86HCFneQpLWQEHsUt9/ipivsiGwq40zivBGlT5BJ5PgDcJUnEqsFOXJksKAy+u4WoUVxbzWZfBGCJDzS94igwB5an4SaKXgpW1MJERa2cPQ55gx6PD0zV6ZfDijWoTX94unevbd5FOODHIIWKIxCH0iLDzJVxYENbta8ATOpuBMbuynSaSGyC4iO7trRXE8oHZLU/QN8tmJ8vMQiNghgsOWVUQC621dxoMS4og3IPgmkx55v+xsgk+x7ZTPhS+yfxTNL7oSvNgBkUokjlfb2B2hg+hkHQMPQdzjlYU9wNkRhoW0jCcs/EXsYGse/InmZj9tzlspIskcQ/CAdRGIJIQxH8RhDCEy3NU+vt3pM0W1eYjghSvClnufN4WkNCFAefDBnrwV5I0i93+1BYjYkRzsiRMnurc5RO3xH96a8saAB+HscdQnRYogEXvKI66JMvMlN7bcrDXOSPC2/w6Q4G0/c11RBERABERABERABESgjQQkeNsIW5cSAREQAREQAREQARFoPwEJ3vYz1xVFQAREQAREQAREQATaSECCt42wdSkREAEREAEREAEREIH2E5DgbT/z0l2RLXtYBNBIYYN4FpSplJuAfKPc9s3bO/lFXnLFPU42L67tytJyCd6yWLIb+1G5RU9IU8q47VdIv1OrI99IzeJh/ZVfhHEqUy3ZvEzWLGZfJHiLaTe1WgREQAREQAREQAREIJCABG8gKFUTAREQAREQAREQAREoJgEJ3mLaTa0WAREQAREQAREQAREIJCDBGwhK1URABERABERABERABIpJQIK3mHZTq0VABERABERABERABAIJSPAGglI1ERABERABERABERCBYhKQ4C2m3dRqERABERABERABERCBQAISvIGgVE0EREAEREAEREAERKCYBCR4i2k3tVoEREAEREAEREAERCCQgARvIChVEwEREAEREAEREAERKCYBCd5i2k2tFgEREAEREAEREAERCCQgwRsIStVEQAREQAREQAREQASKSUCCt5h2U6tFQAREQAREQAREQAQCCUjwBoJSNREQAREQAREQAREQgWISkOAtpt3UahEQAREQAREQAREQgUACEryBoFRNBERABERABERABESgmAQkeItpN7VaBERABERABERABEQgkIAEbyAoVRMBERABERABERABESgmAQneYtpNrRYBERABERABERABEQgkIMEbCErVREAEREAEREAEREAEiklAgreYdlOrRUAEREAEREAEREAEAglI8AaCUjUREAEREAEREAEREIFiEpDgLabd1GoREAEREAEREAEREIFAAhK8gaBUTQREQAREQAREQAREoJgEJHiLaTe1WgREQAREQAREQAREIJCABG8gKFUTAREQAREQAREQAREoJgEJ3mLaTa0WAREQAREQAREQAREIJCDBGwhK1URABERABERABERABIpJQIK3mHZTq0VABERABERABERABAIJSPAGglI1ERABERABERABERCBYhKQ4C2m3dRqERABERABERABERCBQAISvIGgVE0EREAEREAEREAERKCYBCR4i2k3tVoEREAEREAEREAERCCQgARvIChVEwEREAEREAEREAERKCYBCd5i2k2tFgEREAEREAEREAERCCTwf+7gAARVfcjWAAAAAElFTkSuQmCC\" width=\"400\">"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_____\n",
      "model: conv_1d \n",
      "\n",
      "mean best epoc: [15.]\n",
      "mean training time: [59.80732369]\n",
      "STATS:\n",
      "acc           0.964292\n",
      "auc           0.993495\n",
      "loss          0.074132\n",
      "lr            0.000125\n",
      "recall        1.000000\n",
      "val_acc       0.915024\n",
      "val_auc       0.993533\n",
      "val_loss      0.397048\n",
      "val_recall    1.000000\n",
      "dtype: float64\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/javascript": [
       "/* Put everything inside the global mpl namespace */\n",
       "window.mpl = {};\n",
       "\n",
       "\n",
       "mpl.get_websocket_type = function() {\n",
       "    if (typeof(WebSocket) !== 'undefined') {\n",
       "        return WebSocket;\n",
       "    } else if (typeof(MozWebSocket) !== 'undefined') {\n",
       "        return MozWebSocket;\n",
       "    } else {\n",
       "        alert('Your browser does not have WebSocket support.' +\n",
       "              'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
       "              'Firefox 4 and 5 are also supported but you ' +\n",
       "              'have to enable WebSockets in about:config.');\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
       "    this.id = figure_id;\n",
       "\n",
       "    this.ws = websocket;\n",
       "\n",
       "    this.supports_binary = (this.ws.binaryType != undefined);\n",
       "\n",
       "    if (!this.supports_binary) {\n",
       "        var warnings = document.getElementById(\"mpl-warnings\");\n",
       "        if (warnings) {\n",
       "            warnings.style.display = 'block';\n",
       "            warnings.textContent = (\n",
       "                \"This browser does not support binary websocket messages. \" +\n",
       "                    \"Performance may be slow.\");\n",
       "        }\n",
       "    }\n",
       "\n",
       "    this.imageObj = new Image();\n",
       "\n",
       "    this.context = undefined;\n",
       "    this.message = undefined;\n",
       "    this.canvas = undefined;\n",
       "    this.rubberband_canvas = undefined;\n",
       "    this.rubberband_context = undefined;\n",
       "    this.format_dropdown = undefined;\n",
       "\n",
       "    this.image_mode = 'full';\n",
       "\n",
       "    this.root = $('<div/>');\n",
       "    this._root_extra_style(this.root)\n",
       "    this.root.attr('style', 'display: inline-block');\n",
       "\n",
       "    $(parent_element).append(this.root);\n",
       "\n",
       "    this._init_header(this);\n",
       "    this._init_canvas(this);\n",
       "    this._init_toolbar(this);\n",
       "\n",
       "    var fig = this;\n",
       "\n",
       "    this.waiting = false;\n",
       "\n",
       "    this.ws.onopen =  function () {\n",
       "            fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
       "            fig.send_message(\"send_image_mode\", {});\n",
       "            if (mpl.ratio != 1) {\n",
       "                fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
       "            }\n",
       "            fig.send_message(\"refresh\", {});\n",
       "        }\n",
       "\n",
       "    this.imageObj.onload = function() {\n",
       "            if (fig.image_mode == 'full') {\n",
       "                // Full images could contain transparency (where diff images\n",
       "                // almost always do), so we need to clear the canvas so that\n",
       "                // there is no ghosting.\n",
       "                fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
       "            }\n",
       "            fig.context.drawImage(fig.imageObj, 0, 0);\n",
       "        };\n",
       "\n",
       "    this.imageObj.onunload = function() {\n",
       "        fig.ws.close();\n",
       "    }\n",
       "\n",
       "    this.ws.onmessage = this._make_on_message_function(this);\n",
       "\n",
       "    this.ondownload = ondownload;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_header = function() {\n",
       "    var titlebar = $(\n",
       "        '<div class=\"ui-dialog-titlebar ui-widget-header ui-corner-all ' +\n",
       "        'ui-helper-clearfix\"/>');\n",
       "    var titletext = $(\n",
       "        '<div class=\"ui-dialog-title\" style=\"width: 100%; ' +\n",
       "        'text-align: center; padding: 3px;\"/>');\n",
       "    titlebar.append(titletext)\n",
       "    this.root.append(titlebar);\n",
       "    this.header = titletext[0];\n",
       "}\n",
       "\n",
       "\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_canvas = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var canvas_div = $('<div/>');\n",
       "\n",
       "    canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
       "\n",
       "    function canvas_keyboard_event(event) {\n",
       "        return fig.key_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    canvas_div.keydown('key_press', canvas_keyboard_event);\n",
       "    canvas_div.keyup('key_release', canvas_keyboard_event);\n",
       "    this.canvas_div = canvas_div\n",
       "    this._canvas_extra_style(canvas_div)\n",
       "    this.root.append(canvas_div);\n",
       "\n",
       "    var canvas = $('<canvas/>');\n",
       "    canvas.addClass('mpl-canvas');\n",
       "    canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
       "\n",
       "    this.canvas = canvas[0];\n",
       "    this.context = canvas[0].getContext(\"2d\");\n",
       "\n",
       "    var backingStore = this.context.backingStorePixelRatio ||\n",
       "\tthis.context.webkitBackingStorePixelRatio ||\n",
       "\tthis.context.mozBackingStorePixelRatio ||\n",
       "\tthis.context.msBackingStorePixelRatio ||\n",
       "\tthis.context.oBackingStorePixelRatio ||\n",
       "\tthis.context.backingStorePixelRatio || 1;\n",
       "\n",
       "    mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
       "\n",
       "    var rubberband = $('<canvas/>');\n",
       "    rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
       "\n",
       "    var pass_mouse_events = true;\n",
       "\n",
       "    canvas_div.resizable({\n",
       "        start: function(event, ui) {\n",
       "            pass_mouse_events = false;\n",
       "        },\n",
       "        resize: function(event, ui) {\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "        stop: function(event, ui) {\n",
       "            pass_mouse_events = true;\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "    });\n",
       "\n",
       "    function mouse_event_fn(event) {\n",
       "        if (pass_mouse_events)\n",
       "            return fig.mouse_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    rubberband.mousedown('button_press', mouse_event_fn);\n",
       "    rubberband.mouseup('button_release', mouse_event_fn);\n",
       "    // Throttle sequential mouse events to 1 every 20ms.\n",
       "    rubberband.mousemove('motion_notify', mouse_event_fn);\n",
       "\n",
       "    rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
       "    rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
       "\n",
       "    canvas_div.on(\"wheel\", function (event) {\n",
       "        event = event.originalEvent;\n",
       "        event['data'] = 'scroll'\n",
       "        if (event.deltaY < 0) {\n",
       "            event.step = 1;\n",
       "        } else {\n",
       "            event.step = -1;\n",
       "        }\n",
       "        mouse_event_fn(event);\n",
       "    });\n",
       "\n",
       "    canvas_div.append(canvas);\n",
       "    canvas_div.append(rubberband);\n",
       "\n",
       "    this.rubberband = rubberband;\n",
       "    this.rubberband_canvas = rubberband[0];\n",
       "    this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
       "    this.rubberband_context.strokeStyle = \"#000000\";\n",
       "\n",
       "    this._resize_canvas = function(width, height) {\n",
       "        // Keep the size of the canvas, canvas container, and rubber band\n",
       "        // canvas in synch.\n",
       "        canvas_div.css('width', width)\n",
       "        canvas_div.css('height', height)\n",
       "\n",
       "        canvas.attr('width', width * mpl.ratio);\n",
       "        canvas.attr('height', height * mpl.ratio);\n",
       "        canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
       "\n",
       "        rubberband.attr('width', width);\n",
       "        rubberband.attr('height', height);\n",
       "    }\n",
       "\n",
       "    // Set the figure to an initial 600x600px, this will subsequently be updated\n",
       "    // upon first draw.\n",
       "    this._resize_canvas(600, 600);\n",
       "\n",
       "    // Disable right mouse context menu.\n",
       "    $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
       "        return false;\n",
       "    });\n",
       "\n",
       "    function set_focus () {\n",
       "        canvas.focus();\n",
       "        canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    window.setTimeout(set_focus, 100);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>')\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items) {\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) {\n",
       "            // put a spacer in here.\n",
       "            continue;\n",
       "        }\n",
       "        var button = $('<button/>');\n",
       "        button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
       "                        'ui-button-icon-only');\n",
       "        button.attr('role', 'button');\n",
       "        button.attr('aria-disabled', 'false');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "\n",
       "        var icon_img = $('<span/>');\n",
       "        icon_img.addClass('ui-button-icon-primary ui-icon');\n",
       "        icon_img.addClass(image);\n",
       "        icon_img.addClass('ui-corner-all');\n",
       "\n",
       "        var tooltip_span = $('<span/>');\n",
       "        tooltip_span.addClass('ui-button-text');\n",
       "        tooltip_span.html(tooltip);\n",
       "\n",
       "        button.append(icon_img);\n",
       "        button.append(tooltip_span);\n",
       "\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    var fmt_picker_span = $('<span/>');\n",
       "\n",
       "    var fmt_picker = $('<select/>');\n",
       "    fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
       "    fmt_picker_span.append(fmt_picker);\n",
       "    nav_element.append(fmt_picker_span);\n",
       "    this.format_dropdown = fmt_picker[0];\n",
       "\n",
       "    for (var ind in mpl.extensions) {\n",
       "        var fmt = mpl.extensions[ind];\n",
       "        var option = $(\n",
       "            '<option/>', {selected: fmt === mpl.default_extension}).html(fmt);\n",
       "        fmt_picker.append(option)\n",
       "    }\n",
       "\n",
       "    // Add hover states to the ui-buttons\n",
       "    $( \".ui-button\" ).hover(\n",
       "        function() { $(this).addClass(\"ui-state-hover\");},\n",
       "        function() { $(this).removeClass(\"ui-state-hover\");}\n",
       "    );\n",
       "\n",
       "    var status_bar = $('<span class=\"mpl-message\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
       "    // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
       "    // which will in turn request a refresh of the image.\n",
       "    this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_message = function(type, properties) {\n",
       "    properties['type'] = type;\n",
       "    properties['figure_id'] = this.id;\n",
       "    this.ws.send(JSON.stringify(properties));\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_draw_message = function() {\n",
       "    if (!this.waiting) {\n",
       "        this.waiting = true;\n",
       "        this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
       "    }\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    var format_dropdown = fig.format_dropdown;\n",
       "    var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
       "    fig.ondownload(fig, format);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
       "    var size = msg['size'];\n",
       "    if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
       "        fig._resize_canvas(size[0], size[1]);\n",
       "        fig.send_message(\"refresh\", {});\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
       "    var x0 = msg['x0'] / mpl.ratio;\n",
       "    var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
       "    var x1 = msg['x1'] / mpl.ratio;\n",
       "    var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
       "    x0 = Math.floor(x0) + 0.5;\n",
       "    y0 = Math.floor(y0) + 0.5;\n",
       "    x1 = Math.floor(x1) + 0.5;\n",
       "    y1 = Math.floor(y1) + 0.5;\n",
       "    var min_x = Math.min(x0, x1);\n",
       "    var min_y = Math.min(y0, y1);\n",
       "    var width = Math.abs(x1 - x0);\n",
       "    var height = Math.abs(y1 - y0);\n",
       "\n",
       "    fig.rubberband_context.clearRect(\n",
       "        0, 0, fig.canvas.width, fig.canvas.height);\n",
       "\n",
       "    fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
       "    // Updates the figure title.\n",
       "    fig.header.textContent = msg['label'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
       "    var cursor = msg['cursor'];\n",
       "    switch(cursor)\n",
       "    {\n",
       "    case 0:\n",
       "        cursor = 'pointer';\n",
       "        break;\n",
       "    case 1:\n",
       "        cursor = 'default';\n",
       "        break;\n",
       "    case 2:\n",
       "        cursor = 'crosshair';\n",
       "        break;\n",
       "    case 3:\n",
       "        cursor = 'move';\n",
       "        break;\n",
       "    }\n",
       "    fig.rubberband_canvas.style.cursor = cursor;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_message = function(fig, msg) {\n",
       "    fig.message.textContent = msg['message'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
       "    // Request the server to send over a new figure.\n",
       "    fig.send_draw_message();\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
       "    fig.image_mode = msg['mode'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Called whenever the canvas gets updated.\n",
       "    this.send_message(\"ack\", {});\n",
       "}\n",
       "\n",
       "// A function to construct a web socket function for onmessage handling.\n",
       "// Called in the figure constructor.\n",
       "mpl.figure.prototype._make_on_message_function = function(fig) {\n",
       "    return function socket_on_message(evt) {\n",
       "        if (evt.data instanceof Blob) {\n",
       "            /* FIXME: We get \"Resource interpreted as Image but\n",
       "             * transferred with MIME type text/plain:\" errors on\n",
       "             * Chrome.  But how to set the MIME type?  It doesn't seem\n",
       "             * to be part of the websocket stream */\n",
       "            evt.data.type = \"image/png\";\n",
       "\n",
       "            /* Free the memory for the previous frames */\n",
       "            if (fig.imageObj.src) {\n",
       "                (window.URL || window.webkitURL).revokeObjectURL(\n",
       "                    fig.imageObj.src);\n",
       "            }\n",
       "\n",
       "            fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
       "                evt.data);\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "        else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
       "            fig.imageObj.src = evt.data;\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        var msg = JSON.parse(evt.data);\n",
       "        var msg_type = msg['type'];\n",
       "\n",
       "        // Call the  \"handle_{type}\" callback, which takes\n",
       "        // the figure and JSON message as its only arguments.\n",
       "        try {\n",
       "            var callback = fig[\"handle_\" + msg_type];\n",
       "        } catch (e) {\n",
       "            console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        if (callback) {\n",
       "            try {\n",
       "                // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
       "                callback(fig, msg);\n",
       "            } catch (e) {\n",
       "                console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
       "            }\n",
       "        }\n",
       "    };\n",
       "}\n",
       "\n",
       "// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
       "mpl.findpos = function(e) {\n",
       "    //this section is from http://www.quirksmode.org/js/events_properties.html\n",
       "    var targ;\n",
       "    if (!e)\n",
       "        e = window.event;\n",
       "    if (e.target)\n",
       "        targ = e.target;\n",
       "    else if (e.srcElement)\n",
       "        targ = e.srcElement;\n",
       "    if (targ.nodeType == 3) // defeat Safari bug\n",
       "        targ = targ.parentNode;\n",
       "\n",
       "    // jQuery normalizes the pageX and pageY\n",
       "    // pageX,Y are the mouse positions relative to the document\n",
       "    // offset() returns the position of the element relative to the document\n",
       "    var x = e.pageX - $(targ).offset().left;\n",
       "    var y = e.pageY - $(targ).offset().top;\n",
       "\n",
       "    return {\"x\": x, \"y\": y};\n",
       "};\n",
       "\n",
       "/*\n",
       " * return a copy of an object with only non-object keys\n",
       " * we need this to avoid circular references\n",
       " * http://stackoverflow.com/a/24161582/3208463\n",
       " */\n",
       "function simpleKeys (original) {\n",
       "  return Object.keys(original).reduce(function (obj, key) {\n",
       "    if (typeof original[key] !== 'object')\n",
       "        obj[key] = original[key]\n",
       "    return obj;\n",
       "  }, {});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.mouse_event = function(event, name) {\n",
       "    var canvas_pos = mpl.findpos(event)\n",
       "\n",
       "    if (name === 'button_press')\n",
       "    {\n",
       "        this.canvas.focus();\n",
       "        this.canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    var x = canvas_pos.x * mpl.ratio;\n",
       "    var y = canvas_pos.y * mpl.ratio;\n",
       "\n",
       "    this.send_message(name, {x: x, y: y, button: event.button,\n",
       "                             step: event.step,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "\n",
       "    /* This prevents the web browser from automatically changing to\n",
       "     * the text insertion cursor when the button is pressed.  We want\n",
       "     * to control all of the cursor setting manually through the\n",
       "     * 'cursor' event from matplotlib */\n",
       "    event.preventDefault();\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    // Handle any extra behaviour associated with a key event\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.key_event = function(event, name) {\n",
       "\n",
       "    // Prevent repeat events\n",
       "    if (name == 'key_press')\n",
       "    {\n",
       "        if (event.which === this._key)\n",
       "            return;\n",
       "        else\n",
       "            this._key = event.which;\n",
       "    }\n",
       "    if (name == 'key_release')\n",
       "        this._key = null;\n",
       "\n",
       "    var value = '';\n",
       "    if (event.ctrlKey && event.which != 17)\n",
       "        value += \"ctrl+\";\n",
       "    if (event.altKey && event.which != 18)\n",
       "        value += \"alt+\";\n",
       "    if (event.shiftKey && event.which != 16)\n",
       "        value += \"shift+\";\n",
       "\n",
       "    value += 'k';\n",
       "    value += event.which.toString();\n",
       "\n",
       "    this._key_event_extra(event, name);\n",
       "\n",
       "    this.send_message(name, {key: value,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
       "    if (name == 'download') {\n",
       "        this.handle_save(this, null);\n",
       "    } else {\n",
       "        this.send_message(\"toolbar_button\", {name: name});\n",
       "    }\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
       "    this.message.textContent = tooltip;\n",
       "};\n",
       "mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
       "\n",
       "mpl.extensions = [\"eps\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\"];\n",
       "\n",
       "mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
       "    // Create a \"websocket\"-like object which calls the given IPython comm\n",
       "    // object with the appropriate methods. Currently this is a non binary\n",
       "    // socket, so there is still some room for performance tuning.\n",
       "    var ws = {};\n",
       "\n",
       "    ws.close = function() {\n",
       "        comm.close()\n",
       "    };\n",
       "    ws.send = function(m) {\n",
       "        //console.log('sending', m);\n",
       "        comm.send(m);\n",
       "    };\n",
       "    // Register the callback with on_msg.\n",
       "    comm.on_msg(function(msg) {\n",
       "        //console.log('receiving', msg['content']['data'], msg);\n",
       "        // Pass the mpl event to the overridden (by mpl) onmessage function.\n",
       "        ws.onmessage(msg['content']['data'])\n",
       "    });\n",
       "    return ws;\n",
       "}\n",
       "\n",
       "mpl.mpl_figure_comm = function(comm, msg) {\n",
       "    // This is the function which gets called when the mpl process\n",
       "    // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
       "\n",
       "    var id = msg.content.data.id;\n",
       "    // Get hold of the div created by the display call when the Comm\n",
       "    // socket was opened in Python.\n",
       "    var element = $(\"#\" + id);\n",
       "    var ws_proxy = comm_websocket_adapter(comm)\n",
       "\n",
       "    function ondownload(figure, format) {\n",
       "        window.open(figure.imageObj.src);\n",
       "    }\n",
       "\n",
       "    var fig = new mpl.figure(id, ws_proxy,\n",
       "                           ondownload,\n",
       "                           element.get(0));\n",
       "\n",
       "    // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
       "    // web socket which is closed, not our websocket->open comm proxy.\n",
       "    ws_proxy.onopen();\n",
       "\n",
       "    fig.parent_element = element.get(0);\n",
       "    fig.cell_info = mpl.find_output_cell(\"<div id='\" + id + \"'></div>\");\n",
       "    if (!fig.cell_info) {\n",
       "        console.error(\"Failed to find cell for figure\", id, fig);\n",
       "        return;\n",
       "    }\n",
       "\n",
       "    var output_index = fig.cell_info[2]\n",
       "    var cell = fig.cell_info[0];\n",
       "\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.handle_close = function(fig, msg) {\n",
       "    var width = fig.canvas.width/mpl.ratio\n",
       "    fig.root.unbind('remove')\n",
       "\n",
       "    // Update the output cell to use the data from the current canvas.\n",
       "    fig.push_to_output();\n",
       "    var dataURL = fig.canvas.toDataURL();\n",
       "    // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
       "    // the notebook keyboard shortcuts fail.\n",
       "    IPython.keyboard_manager.enable()\n",
       "    $(fig.parent_element).html('<img src=\"' + dataURL + '\" width=\"' + width + '\">');\n",
       "    fig.close_ws(fig, msg);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.close_ws = function(fig, msg){\n",
       "    fig.send_message('closing', msg);\n",
       "    // fig.ws.close()\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
       "    // Turn the data on the canvas into data in the output cell.\n",
       "    var width = this.canvas.width/mpl.ratio\n",
       "    var dataURL = this.canvas.toDataURL();\n",
       "    this.cell_info[1]['text/html'] = '<img src=\"' + dataURL + '\" width=\"' + width + '\">';\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Tell IPython that the notebook contents must change.\n",
       "    IPython.notebook.set_dirty(true);\n",
       "    this.send_message(\"ack\", {});\n",
       "    var fig = this;\n",
       "    // Wait a second, then push the new image to the DOM so\n",
       "    // that it is saved nicely (might be nice to debounce this).\n",
       "    setTimeout(function () { fig.push_to_output() }, 1000);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>')\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items){\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) { continue; };\n",
       "\n",
       "        var button = $('<button class=\"btn btn-default\" href=\"#\" title=\"' + name + '\"><i class=\"fa ' + image + ' fa-lg\"></i></button>');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    // Add the status bar.\n",
       "    var status_bar = $('<span class=\"mpl-message\" style=\"text-align:right; float: right;\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "\n",
       "    // Add the close button to the window.\n",
       "    var buttongrp = $('<div class=\"btn-group inline pull-right\"></div>');\n",
       "    var button = $('<button class=\"btn btn-mini btn-primary\" href=\"#\" title=\"Stop Interaction\"><i class=\"fa fa-power-off icon-remove icon-large\"></i></button>');\n",
       "    button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
       "    button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
       "    buttongrp.append(button);\n",
       "    var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
       "    titlebar.prepend(buttongrp);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(el){\n",
       "    var fig = this\n",
       "    el.on(\"remove\", function(){\n",
       "\tfig.close_ws(fig, {});\n",
       "    });\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(el){\n",
       "    // this is important to make the div 'focusable\n",
       "    el.attr('tabindex', 0)\n",
       "    // reach out to IPython and tell the keyboard manager to turn it's self\n",
       "    // off when our div gets focus\n",
       "\n",
       "    // location in version 3\n",
       "    if (IPython.notebook.keyboard_manager) {\n",
       "        IPython.notebook.keyboard_manager.register_events(el);\n",
       "    }\n",
       "    else {\n",
       "        // location in version 2\n",
       "        IPython.keyboard_manager.register_events(el);\n",
       "    }\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    var manager = IPython.notebook.keyboard_manager;\n",
       "    if (!manager)\n",
       "        manager = IPython.keyboard_manager;\n",
       "\n",
       "    // Check for shift+enter\n",
       "    if (event.shiftKey && event.which == 13) {\n",
       "        this.canvas_div.blur();\n",
       "        event.shiftKey = false;\n",
       "        // Send a \"J\" for go to next cell\n",
       "        event.which = 74;\n",
       "        event.keyCode = 74;\n",
       "        manager.command_mode();\n",
       "        manager.handle_keydown(event);\n",
       "    }\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    fig.ondownload(fig, null);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.find_output_cell = function(html_output) {\n",
       "    // Return the cell and output element which can be found *uniquely* in the notebook.\n",
       "    // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
       "    // IPython event is triggered only after the cells have been serialised, which for\n",
       "    // our purposes (turning an active figure into a static one), is too late.\n",
       "    var cells = IPython.notebook.get_cells();\n",
       "    var ncells = cells.length;\n",
       "    for (var i=0; i<ncells; i++) {\n",
       "        var cell = cells[i];\n",
       "        if (cell.cell_type === 'code'){\n",
       "            for (var j=0; j<cell.output_area.outputs.length; j++) {\n",
       "                var data = cell.output_area.outputs[j];\n",
       "                if (data.data) {\n",
       "                    // IPython >= 3 moved mimebundle to data attribute of output\n",
       "                    data = data.data;\n",
       "                }\n",
       "                if (data['text/html'] == html_output) {\n",
       "                    return [cell, data, j];\n",
       "                }\n",
       "            }\n",
       "        }\n",
       "    }\n",
       "}\n",
       "\n",
       "// Register the function which deals with the matplotlib target/channel.\n",
       "// The kernel may be null if the page has been refreshed.\n",
       "if (IPython.notebook.kernel != null) {\n",
       "    IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
       "}\n"
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<img src=\"data:image/png;base64,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\" width=\"400\">"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for model_name in trainig_reulsts.keys():\n",
    "    print('_____')\n",
    "    print('model: {} \\n'.format(model_name))\n",
    "    print('mean best epoc: {}'.format(trainig_reulsts[model_name]['stats_best_epoch'].values))\n",
    "    print('mean training time: {}'.format(trainig_reulsts[model_name]['stats_time_takes'].values))\n",
    "    \n",
    "    print('STATS:')\n",
    "    \n",
    "    print(trainig_reulsts[model_name]['stats_best_stats'])\n",
    "    print('\\n')\n",
    "\n",
    "    plt.figure(figsize = (4,4))\n",
    "    stats_for_model = trainig_reulsts[model_name]['trained_models_stats']\n",
    "    stats_for_model[stats_for_model.columns[stats_for_model.columns.str.contains(pat = 'val')]].boxplot()\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.9927354 , 0.99437265, 0.99348999])"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "stats_for_model[['val_auc']].values.ravel()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "/* Put everything inside the global mpl namespace */\n",
       "window.mpl = {};\n",
       "\n",
       "\n",
       "mpl.get_websocket_type = function() {\n",
       "    if (typeof(WebSocket) !== 'undefined') {\n",
       "        return WebSocket;\n",
       "    } else if (typeof(MozWebSocket) !== 'undefined') {\n",
       "        return MozWebSocket;\n",
       "    } else {\n",
       "        alert('Your browser does not have WebSocket support.' +\n",
       "              'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
       "              'Firefox 4 and 5 are also supported but you ' +\n",
       "              'have to enable WebSockets in about:config.');\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
       "    this.id = figure_id;\n",
       "\n",
       "    this.ws = websocket;\n",
       "\n",
       "    this.supports_binary = (this.ws.binaryType != undefined);\n",
       "\n",
       "    if (!this.supports_binary) {\n",
       "        var warnings = document.getElementById(\"mpl-warnings\");\n",
       "        if (warnings) {\n",
       "            warnings.style.display = 'block';\n",
       "            warnings.textContent = (\n",
       "                \"This browser does not support binary websocket messages. \" +\n",
       "                    \"Performance may be slow.\");\n",
       "        }\n",
       "    }\n",
       "\n",
       "    this.imageObj = new Image();\n",
       "\n",
       "    this.context = undefined;\n",
       "    this.message = undefined;\n",
       "    this.canvas = undefined;\n",
       "    this.rubberband_canvas = undefined;\n",
       "    this.rubberband_context = undefined;\n",
       "    this.format_dropdown = undefined;\n",
       "\n",
       "    this.image_mode = 'full';\n",
       "\n",
       "    this.root = $('<div/>');\n",
       "    this._root_extra_style(this.root)\n",
       "    this.root.attr('style', 'display: inline-block');\n",
       "\n",
       "    $(parent_element).append(this.root);\n",
       "\n",
       "    this._init_header(this);\n",
       "    this._init_canvas(this);\n",
       "    this._init_toolbar(this);\n",
       "\n",
       "    var fig = this;\n",
       "\n",
       "    this.waiting = false;\n",
       "\n",
       "    this.ws.onopen =  function () {\n",
       "            fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
       "            fig.send_message(\"send_image_mode\", {});\n",
       "            if (mpl.ratio != 1) {\n",
       "                fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
       "            }\n",
       "            fig.send_message(\"refresh\", {});\n",
       "        }\n",
       "\n",
       "    this.imageObj.onload = function() {\n",
       "            if (fig.image_mode == 'full') {\n",
       "                // Full images could contain transparency (where diff images\n",
       "                // almost always do), so we need to clear the canvas so that\n",
       "                // there is no ghosting.\n",
       "                fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
       "            }\n",
       "            fig.context.drawImage(fig.imageObj, 0, 0);\n",
       "        };\n",
       "\n",
       "    this.imageObj.onunload = function() {\n",
       "        fig.ws.close();\n",
       "    }\n",
       "\n",
       "    this.ws.onmessage = this._make_on_message_function(this);\n",
       "\n",
       "    this.ondownload = ondownload;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_header = function() {\n",
       "    var titlebar = $(\n",
       "        '<div class=\"ui-dialog-titlebar ui-widget-header ui-corner-all ' +\n",
       "        'ui-helper-clearfix\"/>');\n",
       "    var titletext = $(\n",
       "        '<div class=\"ui-dialog-title\" style=\"width: 100%; ' +\n",
       "        'text-align: center; padding: 3px;\"/>');\n",
       "    titlebar.append(titletext)\n",
       "    this.root.append(titlebar);\n",
       "    this.header = titletext[0];\n",
       "}\n",
       "\n",
       "\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_canvas = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var canvas_div = $('<div/>');\n",
       "\n",
       "    canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
       "\n",
       "    function canvas_keyboard_event(event) {\n",
       "        return fig.key_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    canvas_div.keydown('key_press', canvas_keyboard_event);\n",
       "    canvas_div.keyup('key_release', canvas_keyboard_event);\n",
       "    this.canvas_div = canvas_div\n",
       "    this._canvas_extra_style(canvas_div)\n",
       "    this.root.append(canvas_div);\n",
       "\n",
       "    var canvas = $('<canvas/>');\n",
       "    canvas.addClass('mpl-canvas');\n",
       "    canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
       "\n",
       "    this.canvas = canvas[0];\n",
       "    this.context = canvas[0].getContext(\"2d\");\n",
       "\n",
       "    var backingStore = this.context.backingStorePixelRatio ||\n",
       "\tthis.context.webkitBackingStorePixelRatio ||\n",
       "\tthis.context.mozBackingStorePixelRatio ||\n",
       "\tthis.context.msBackingStorePixelRatio ||\n",
       "\tthis.context.oBackingStorePixelRatio ||\n",
       "\tthis.context.backingStorePixelRatio || 1;\n",
       "\n",
       "    mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
       "\n",
       "    var rubberband = $('<canvas/>');\n",
       "    rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
       "\n",
       "    var pass_mouse_events = true;\n",
       "\n",
       "    canvas_div.resizable({\n",
       "        start: function(event, ui) {\n",
       "            pass_mouse_events = false;\n",
       "        },\n",
       "        resize: function(event, ui) {\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "        stop: function(event, ui) {\n",
       "            pass_mouse_events = true;\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "    });\n",
       "\n",
       "    function mouse_event_fn(event) {\n",
       "        if (pass_mouse_events)\n",
       "            return fig.mouse_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    rubberband.mousedown('button_press', mouse_event_fn);\n",
       "    rubberband.mouseup('button_release', mouse_event_fn);\n",
       "    // Throttle sequential mouse events to 1 every 20ms.\n",
       "    rubberband.mousemove('motion_notify', mouse_event_fn);\n",
       "\n",
       "    rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
       "    rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
       "\n",
       "    canvas_div.on(\"wheel\", function (event) {\n",
       "        event = event.originalEvent;\n",
       "        event['data'] = 'scroll'\n",
       "        if (event.deltaY < 0) {\n",
       "            event.step = 1;\n",
       "        } else {\n",
       "            event.step = -1;\n",
       "        }\n",
       "        mouse_event_fn(event);\n",
       "    });\n",
       "\n",
       "    canvas_div.append(canvas);\n",
       "    canvas_div.append(rubberband);\n",
       "\n",
       "    this.rubberband = rubberband;\n",
       "    this.rubberband_canvas = rubberband[0];\n",
       "    this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
       "    this.rubberband_context.strokeStyle = \"#000000\";\n",
       "\n",
       "    this._resize_canvas = function(width, height) {\n",
       "        // Keep the size of the canvas, canvas container, and rubber band\n",
       "        // canvas in synch.\n",
       "        canvas_div.css('width', width)\n",
       "        canvas_div.css('height', height)\n",
       "\n",
       "        canvas.attr('width', width * mpl.ratio);\n",
       "        canvas.attr('height', height * mpl.ratio);\n",
       "        canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
       "\n",
       "        rubberband.attr('width', width);\n",
       "        rubberband.attr('height', height);\n",
       "    }\n",
       "\n",
       "    // Set the figure to an initial 600x600px, this will subsequently be updated\n",
       "    // upon first draw.\n",
       "    this._resize_canvas(600, 600);\n",
       "\n",
       "    // Disable right mouse context menu.\n",
       "    $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
       "        return false;\n",
       "    });\n",
       "\n",
       "    function set_focus () {\n",
       "        canvas.focus();\n",
       "        canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    window.setTimeout(set_focus, 100);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>')\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items) {\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) {\n",
       "            // put a spacer in here.\n",
       "            continue;\n",
       "        }\n",
       "        var button = $('<button/>');\n",
       "        button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
       "                        'ui-button-icon-only');\n",
       "        button.attr('role', 'button');\n",
       "        button.attr('aria-disabled', 'false');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "\n",
       "        var icon_img = $('<span/>');\n",
       "        icon_img.addClass('ui-button-icon-primary ui-icon');\n",
       "        icon_img.addClass(image);\n",
       "        icon_img.addClass('ui-corner-all');\n",
       "\n",
       "        var tooltip_span = $('<span/>');\n",
       "        tooltip_span.addClass('ui-button-text');\n",
       "        tooltip_span.html(tooltip);\n",
       "\n",
       "        button.append(icon_img);\n",
       "        button.append(tooltip_span);\n",
       "\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    var fmt_picker_span = $('<span/>');\n",
       "\n",
       "    var fmt_picker = $('<select/>');\n",
       "    fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
       "    fmt_picker_span.append(fmt_picker);\n",
       "    nav_element.append(fmt_picker_span);\n",
       "    this.format_dropdown = fmt_picker[0];\n",
       "\n",
       "    for (var ind in mpl.extensions) {\n",
       "        var fmt = mpl.extensions[ind];\n",
       "        var option = $(\n",
       "            '<option/>', {selected: fmt === mpl.default_extension}).html(fmt);\n",
       "        fmt_picker.append(option)\n",
       "    }\n",
       "\n",
       "    // Add hover states to the ui-buttons\n",
       "    $( \".ui-button\" ).hover(\n",
       "        function() { $(this).addClass(\"ui-state-hover\");},\n",
       "        function() { $(this).removeClass(\"ui-state-hover\");}\n",
       "    );\n",
       "\n",
       "    var status_bar = $('<span class=\"mpl-message\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
       "    // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
       "    // which will in turn request a refresh of the image.\n",
       "    this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_message = function(type, properties) {\n",
       "    properties['type'] = type;\n",
       "    properties['figure_id'] = this.id;\n",
       "    this.ws.send(JSON.stringify(properties));\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_draw_message = function() {\n",
       "    if (!this.waiting) {\n",
       "        this.waiting = true;\n",
       "        this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
       "    }\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    var format_dropdown = fig.format_dropdown;\n",
       "    var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
       "    fig.ondownload(fig, format);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
       "    var size = msg['size'];\n",
       "    if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
       "        fig._resize_canvas(size[0], size[1]);\n",
       "        fig.send_message(\"refresh\", {});\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
       "    var x0 = msg['x0'] / mpl.ratio;\n",
       "    var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
       "    var x1 = msg['x1'] / mpl.ratio;\n",
       "    var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
       "    x0 = Math.floor(x0) + 0.5;\n",
       "    y0 = Math.floor(y0) + 0.5;\n",
       "    x1 = Math.floor(x1) + 0.5;\n",
       "    y1 = Math.floor(y1) + 0.5;\n",
       "    var min_x = Math.min(x0, x1);\n",
       "    var min_y = Math.min(y0, y1);\n",
       "    var width = Math.abs(x1 - x0);\n",
       "    var height = Math.abs(y1 - y0);\n",
       "\n",
       "    fig.rubberband_context.clearRect(\n",
       "        0, 0, fig.canvas.width, fig.canvas.height);\n",
       "\n",
       "    fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
       "    // Updates the figure title.\n",
       "    fig.header.textContent = msg['label'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
       "    var cursor = msg['cursor'];\n",
       "    switch(cursor)\n",
       "    {\n",
       "    case 0:\n",
       "        cursor = 'pointer';\n",
       "        break;\n",
       "    case 1:\n",
       "        cursor = 'default';\n",
       "        break;\n",
       "    case 2:\n",
       "        cursor = 'crosshair';\n",
       "        break;\n",
       "    case 3:\n",
       "        cursor = 'move';\n",
       "        break;\n",
       "    }\n",
       "    fig.rubberband_canvas.style.cursor = cursor;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_message = function(fig, msg) {\n",
       "    fig.message.textContent = msg['message'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
       "    // Request the server to send over a new figure.\n",
       "    fig.send_draw_message();\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
       "    fig.image_mode = msg['mode'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Called whenever the canvas gets updated.\n",
       "    this.send_message(\"ack\", {});\n",
       "}\n",
       "\n",
       "// A function to construct a web socket function for onmessage handling.\n",
       "// Called in the figure constructor.\n",
       "mpl.figure.prototype._make_on_message_function = function(fig) {\n",
       "    return function socket_on_message(evt) {\n",
       "        if (evt.data instanceof Blob) {\n",
       "            /* FIXME: We get \"Resource interpreted as Image but\n",
       "             * transferred with MIME type text/plain:\" errors on\n",
       "             * Chrome.  But how to set the MIME type?  It doesn't seem\n",
       "             * to be part of the websocket stream */\n",
       "            evt.data.type = \"image/png\";\n",
       "\n",
       "            /* Free the memory for the previous frames */\n",
       "            if (fig.imageObj.src) {\n",
       "                (window.URL || window.webkitURL).revokeObjectURL(\n",
       "                    fig.imageObj.src);\n",
       "            }\n",
       "\n",
       "            fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
       "                evt.data);\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "        else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
       "            fig.imageObj.src = evt.data;\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        var msg = JSON.parse(evt.data);\n",
       "        var msg_type = msg['type'];\n",
       "\n",
       "        // Call the  \"handle_{type}\" callback, which takes\n",
       "        // the figure and JSON message as its only arguments.\n",
       "        try {\n",
       "            var callback = fig[\"handle_\" + msg_type];\n",
       "        } catch (e) {\n",
       "            console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        if (callback) {\n",
       "            try {\n",
       "                // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
       "                callback(fig, msg);\n",
       "            } catch (e) {\n",
       "                console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
       "            }\n",
       "        }\n",
       "    };\n",
       "}\n",
       "\n",
       "// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
       "mpl.findpos = function(e) {\n",
       "    //this section is from http://www.quirksmode.org/js/events_properties.html\n",
       "    var targ;\n",
       "    if (!e)\n",
       "        e = window.event;\n",
       "    if (e.target)\n",
       "        targ = e.target;\n",
       "    else if (e.srcElement)\n",
       "        targ = e.srcElement;\n",
       "    if (targ.nodeType == 3) // defeat Safari bug\n",
       "        targ = targ.parentNode;\n",
       "\n",
       "    // jQuery normalizes the pageX and pageY\n",
       "    // pageX,Y are the mouse positions relative to the document\n",
       "    // offset() returns the position of the element relative to the document\n",
       "    var x = e.pageX - $(targ).offset().left;\n",
       "    var y = e.pageY - $(targ).offset().top;\n",
       "\n",
       "    return {\"x\": x, \"y\": y};\n",
       "};\n",
       "\n",
       "/*\n",
       " * return a copy of an object with only non-object keys\n",
       " * we need this to avoid circular references\n",
       " * http://stackoverflow.com/a/24161582/3208463\n",
       " */\n",
       "function simpleKeys (original) {\n",
       "  return Object.keys(original).reduce(function (obj, key) {\n",
       "    if (typeof original[key] !== 'object')\n",
       "        obj[key] = original[key]\n",
       "    return obj;\n",
       "  }, {});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.mouse_event = function(event, name) {\n",
       "    var canvas_pos = mpl.findpos(event)\n",
       "\n",
       "    if (name === 'button_press')\n",
       "    {\n",
       "        this.canvas.focus();\n",
       "        this.canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    var x = canvas_pos.x * mpl.ratio;\n",
       "    var y = canvas_pos.y * mpl.ratio;\n",
       "\n",
       "    this.send_message(name, {x: x, y: y, button: event.button,\n",
       "                             step: event.step,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "\n",
       "    /* This prevents the web browser from automatically changing to\n",
       "     * the text insertion cursor when the button is pressed.  We want\n",
       "     * to control all of the cursor setting manually through the\n",
       "     * 'cursor' event from matplotlib */\n",
       "    event.preventDefault();\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    // Handle any extra behaviour associated with a key event\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.key_event = function(event, name) {\n",
       "\n",
       "    // Prevent repeat events\n",
       "    if (name == 'key_press')\n",
       "    {\n",
       "        if (event.which === this._key)\n",
       "            return;\n",
       "        else\n",
       "            this._key = event.which;\n",
       "    }\n",
       "    if (name == 'key_release')\n",
       "        this._key = null;\n",
       "\n",
       "    var value = '';\n",
       "    if (event.ctrlKey && event.which != 17)\n",
       "        value += \"ctrl+\";\n",
       "    if (event.altKey && event.which != 18)\n",
       "        value += \"alt+\";\n",
       "    if (event.shiftKey && event.which != 16)\n",
       "        value += \"shift+\";\n",
       "\n",
       "    value += 'k';\n",
       "    value += event.which.toString();\n",
       "\n",
       "    this._key_event_extra(event, name);\n",
       "\n",
       "    this.send_message(name, {key: value,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
       "    if (name == 'download') {\n",
       "        this.handle_save(this, null);\n",
       "    } else {\n",
       "        this.send_message(\"toolbar_button\", {name: name});\n",
       "    }\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
       "    this.message.textContent = tooltip;\n",
       "};\n",
       "mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
       "\n",
       "mpl.extensions = [\"eps\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\"];\n",
       "\n",
       "mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
       "    // Create a \"websocket\"-like object which calls the given IPython comm\n",
       "    // object with the appropriate methods. Currently this is a non binary\n",
       "    // socket, so there is still some room for performance tuning.\n",
       "    var ws = {};\n",
       "\n",
       "    ws.close = function() {\n",
       "        comm.close()\n",
       "    };\n",
       "    ws.send = function(m) {\n",
       "        //console.log('sending', m);\n",
       "        comm.send(m);\n",
       "    };\n",
       "    // Register the callback with on_msg.\n",
       "    comm.on_msg(function(msg) {\n",
       "        //console.log('receiving', msg['content']['data'], msg);\n",
       "        // Pass the mpl event to the overridden (by mpl) onmessage function.\n",
       "        ws.onmessage(msg['content']['data'])\n",
       "    });\n",
       "    return ws;\n",
       "}\n",
       "\n",
       "mpl.mpl_figure_comm = function(comm, msg) {\n",
       "    // This is the function which gets called when the mpl process\n",
       "    // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
       "\n",
       "    var id = msg.content.data.id;\n",
       "    // Get hold of the div created by the display call when the Comm\n",
       "    // socket was opened in Python.\n",
       "    var element = $(\"#\" + id);\n",
       "    var ws_proxy = comm_websocket_adapter(comm)\n",
       "\n",
       "    function ondownload(figure, format) {\n",
       "        window.open(figure.imageObj.src);\n",
       "    }\n",
       "\n",
       "    var fig = new mpl.figure(id, ws_proxy,\n",
       "                           ondownload,\n",
       "                           element.get(0));\n",
       "\n",
       "    // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
       "    // web socket which is closed, not our websocket->open comm proxy.\n",
       "    ws_proxy.onopen();\n",
       "\n",
       "    fig.parent_element = element.get(0);\n",
       "    fig.cell_info = mpl.find_output_cell(\"<div id='\" + id + \"'></div>\");\n",
       "    if (!fig.cell_info) {\n",
       "        console.error(\"Failed to find cell for figure\", id, fig);\n",
       "        return;\n",
       "    }\n",
       "\n",
       "    var output_index = fig.cell_info[2]\n",
       "    var cell = fig.cell_info[0];\n",
       "\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.handle_close = function(fig, msg) {\n",
       "    var width = fig.canvas.width/mpl.ratio\n",
       "    fig.root.unbind('remove')\n",
       "\n",
       "    // Update the output cell to use the data from the current canvas.\n",
       "    fig.push_to_output();\n",
       "    var dataURL = fig.canvas.toDataURL();\n",
       "    // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
       "    // the notebook keyboard shortcuts fail.\n",
       "    IPython.keyboard_manager.enable()\n",
       "    $(fig.parent_element).html('<img src=\"' + dataURL + '\" width=\"' + width + '\">');\n",
       "    fig.close_ws(fig, msg);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.close_ws = function(fig, msg){\n",
       "    fig.send_message('closing', msg);\n",
       "    // fig.ws.close()\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
       "    // Turn the data on the canvas into data in the output cell.\n",
       "    var width = this.canvas.width/mpl.ratio\n",
       "    var dataURL = this.canvas.toDataURL();\n",
       "    this.cell_info[1]['text/html'] = '<img src=\"' + dataURL + '\" width=\"' + width + '\">';\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Tell IPython that the notebook contents must change.\n",
       "    IPython.notebook.set_dirty(true);\n",
       "    this.send_message(\"ack\", {});\n",
       "    var fig = this;\n",
       "    // Wait a second, then push the new image to the DOM so\n",
       "    // that it is saved nicely (might be nice to debounce this).\n",
       "    setTimeout(function () { fig.push_to_output() }, 1000);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>')\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items){\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) { continue; };\n",
       "\n",
       "        var button = $('<button class=\"btn btn-default\" href=\"#\" title=\"' + name + '\"><i class=\"fa ' + image + ' fa-lg\"></i></button>');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    // Add the status bar.\n",
       "    var status_bar = $('<span class=\"mpl-message\" style=\"text-align:right; float: right;\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "\n",
       "    // Add the close button to the window.\n",
       "    var buttongrp = $('<div class=\"btn-group inline pull-right\"></div>');\n",
       "    var button = $('<button class=\"btn btn-mini btn-primary\" href=\"#\" title=\"Stop Interaction\"><i class=\"fa fa-power-off icon-remove icon-large\"></i></button>');\n",
       "    button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
       "    button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
       "    buttongrp.append(button);\n",
       "    var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
       "    titlebar.prepend(buttongrp);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(el){\n",
       "    var fig = this\n",
       "    el.on(\"remove\", function(){\n",
       "\tfig.close_ws(fig, {});\n",
       "    });\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(el){\n",
       "    // this is important to make the div 'focusable\n",
       "    el.attr('tabindex', 0)\n",
       "    // reach out to IPython and tell the keyboard manager to turn it's self\n",
       "    // off when our div gets focus\n",
       "\n",
       "    // location in version 3\n",
       "    if (IPython.notebook.keyboard_manager) {\n",
       "        IPython.notebook.keyboard_manager.register_events(el);\n",
       "    }\n",
       "    else {\n",
       "        // location in version 2\n",
       "        IPython.keyboard_manager.register_events(el);\n",
       "    }\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    var manager = IPython.notebook.keyboard_manager;\n",
       "    if (!manager)\n",
       "        manager = IPython.keyboard_manager;\n",
       "\n",
       "    // Check for shift+enter\n",
       "    if (event.shiftKey && event.which == 13) {\n",
       "        this.canvas_div.blur();\n",
       "        event.shiftKey = false;\n",
       "        // Send a \"J\" for go to next cell\n",
       "        event.which = 74;\n",
       "        event.keyCode = 74;\n",
       "        manager.command_mode();\n",
       "        manager.handle_keydown(event);\n",
       "    }\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    fig.ondownload(fig, null);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.find_output_cell = function(html_output) {\n",
       "    // Return the cell and output element which can be found *uniquely* in the notebook.\n",
       "    // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
       "    // IPython event is triggered only after the cells have been serialised, which for\n",
       "    // our purposes (turning an active figure into a static one), is too late.\n",
       "    var cells = IPython.notebook.get_cells();\n",
       "    var ncells = cells.length;\n",
       "    for (var i=0; i<ncells; i++) {\n",
       "        var cell = cells[i];\n",
       "        if (cell.cell_type === 'code'){\n",
       "            for (var j=0; j<cell.output_area.outputs.length; j++) {\n",
       "                var data = cell.output_area.outputs[j];\n",
       "                if (data.data) {\n",
       "                    // IPython >= 3 moved mimebundle to data attribute of output\n",
       "                    data = data.data;\n",
       "                }\n",
       "                if (data['text/html'] == html_output) {\n",
       "                    return [cell, data, j];\n",
       "                }\n",
       "            }\n",
       "        }\n",
       "    }\n",
       "}\n",
       "\n",
       "// Register the function which deals with the matplotlib target/channel.\n",
       "// The kernel may be null if the page has been refreshed.\n",
       "if (IPython.notebook.kernel != null) {\n",
       "    IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
       "}\n"
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<img src=\"data:image/png;base64,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\" width=\"400\">"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val_auc\n",
      "LSTM\t0.9592231799355174\n",
      "dense_2\t0.9934404347587437\n",
      "dense\t0.992079216081108\n",
      "ens\t0.9897511921689647\n",
      "dense_fc\t0.9885121683617338\n",
      "conv_1d\t0.9935326797200957\n",
      "val_acc\n",
      "LSTM\t0.8527079100055595\n",
      "dense_2\t0.9268071690087628\n",
      "dense\t0.915250398365516\n",
      "ens\t0.9052798538663991\n",
      "dense_fc\t0.9079990991383432\n",
      "conv_1d\t0.9150237890832851\n",
      "val_recall\n",
      "LSTM\t0.9839301518969273\n",
      "dense_2\t1.0\n",
      "dense\t1.0\n",
      "ens\t0.9965187410040829\n",
      "dense_fc\t1.0\n",
      "conv_1d\t1.0\n"
     ]
    }
   ],
   "source": [
    "plt.figure(figsize = (4,4))\n",
    "model_names = trainig_reulsts.keys()\n",
    "for i, metric in enumerate(['val_auc', 'val_acc', 'val_recall']):\n",
    "    plt.subplot(3,1,i+1)\n",
    "    plt.title(metric)\n",
    "    which = 'trained_models_stats'\n",
    "    plot_metrics = [trainig_reulsts[model_name][which][metric] for model_name in model_names]\n",
    "    plt.boxplot(plot_metrics, labels= model_names)\n",
    "    plt.tight_layout()    \n",
    "    print(metric)\n",
    "    for model_name in model_names:\n",
    "        print(model_name + '\\t' + str(trainig_reulsts[model_name]['stats_best_stats'][metric]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_name = 'dense'\n",
    "trained_models = []\n",
    "\n",
    "print_summary_only_once = True\n",
    "best_model_epoch = []\n",
    "trained_models_stats = []\n",
    "trained_models_time_taken = []\n",
    "\n",
    "\n",
    "for i in range(N_train):\n",
    "    tic = time()\n",
    "    model_name = model_name + '_' + str(i)\n",
    "    tensorboard = TensorBoard(log_dir=\"logs/{}\".format(model_name + '_' + str(time())))\n",
    "    callbacks_model = callbacks + [tensorboard]\n",
    "\n",
    "    # generate model\n",
    "    model_input, model_output , _ = dense_model_generator(n_timesteps, n_features)\n",
    "    model = Model(model_input, model_output, name = model_name)\n",
    "    \n",
    "    #compile model\n",
    "    model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy', auc_roc, recall])\n",
    "    if print_summary_only_once:\n",
    "        model.summary()\n",
    "        print_summary_only_once = False\n",
    "    \n",
    "    # train model\n",
    "    model.fit(x_train, \n",
    "              y_train, epochs=epochs, \n",
    "              batch_size=batch_size, \n",
    "              validation_split=validation_split_on_training,\n",
    "              verbose=True,\n",
    "              callbacks = callbacks_model)     \n",
    "    trained_models.append(model)\n",
    "    \n",
    "    # training time\n",
    "    training_time = time()-tic\n",
    "    trained_models_time_taken.append(training_time)\n",
    "    print('training time: {}s'.format(training_time))\n",
    "    \n",
    "    # early stopping epoch\n",
    "    best_epoch = es_cb.stopped_epoch\n",
    "    best_model_epoch.append(best_epoch)\n",
    "    \n",
    "    #append best stat\n",
    "    best_stats = {}\n",
    "    for key in model.history.history.keys():\n",
    "        best_stats[key] = model.history.history[key][best_epoch]\n",
    "    trained_models_stats.append(best_stats)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_name = 'dense'\n",
    "tensorboard = TensorBoard(log_dir=\"logs/{}\".format(model_name + '_' + str(time())))\n",
    "callbacks_dense = callbacks + [tensorboard]\n",
    "\n",
    "#generate model\n",
    "dense_input, dense_output , _ = dense_model_generator(n_timesteps, n_features)\n",
    "model = Model(dense_input, dense_output, name = model_name)\n",
    "\n",
    "#compile model\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "model.summary()\n",
    "model.fit(x_train, \n",
    "          y_train, epochs=epochs, \n",
    "          batch_size=batch_size, \n",
    "          validation_split=validation_split_on_training,\n",
    "          verbose=True,\n",
    "         callbacks = callbacks_dense)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "model_name = 'LSTM'\n",
    "tensorboard = TensorBoard(log_dir=\"logs/{}\".format(model_name + '_' + str(time())))\n",
    "callbacks_lstm = callbacks + [tensorboard]\n",
    "\n",
    "lstm_input, lstm_output, _  = lstm_model_generator(n_timesteps, n_features)\n",
    "\n",
    "model = Model(lstm_input, lstm_output, name= model_name )\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "\n",
    "model.fit(x_train, \n",
    "          y_train, epochs=epochs, \n",
    "          batch_size=batch_size, \n",
    "          validation_split=validation_split_on_training,\n",
    "          verbose=True,\n",
    "         callbacks = callbacks_lstm)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Dense network approach"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "some comments about this architecture:\n",
    "\n",
    "- Note that the unstack dim is the feature dimension"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_name = 'dense'\n",
    "tensorboard = TensorBoard(log_dir=\"logs/{}\".format(model_name + '_' + str(time())))\n",
    "callbacks_dense = callbacks + [tensorboard]\n",
    "\n",
    "#generate model\n",
    "dense_input, dense_output , _ = dense_model_generator(n_timesteps, n_features)\n",
    "model = Model(dense_input, dense_output, name = model_name)\n",
    "\n",
    "#compile model\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "model.summary()\n",
    "model.fit(x_train, \n",
    "          y_train, epochs=epochs, \n",
    "          batch_size=batch_size, \n",
    "          validation_split=validation_split_on_training,\n",
    "          verbose=True,\n",
    "         callbacks = callbacks_dense)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Hybrid ensemble"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_name = 'ens'\n",
    "tensorboard = TensorBoard(log_dir=\"logs/{}\".format(model_name + '_' + str(time())))\n",
    "callbacks_ens = callbacks + [tensorboard]\n",
    "\n",
    "ens_input, ens_output, _  = hibrid_ens_generator(n_timesteps, n_features)\n",
    "\n",
    "\n",
    "model = Model(ens_input, ens_output, name= model_name)\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "model.summary()\n",
    "\n",
    "model.fit(x_train, \n",
    "          y_train, epochs=epochs, \n",
    "          batch_size=batch_size, \n",
    "          validation_split=validation_split_on_training,\n",
    "          verbose=True,\n",
    "         callbacks = callbacks_ens)\n",
    "\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}