Newsgroup Text Analysis – Word Embedding 1.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Newsgroup Text Analysis Exercise\n",
    "\n",
    "### Classification into 6 classes using F Chollets orginal model\n",
    "\n",
    "This is a small exercise in text analysis following https://blog.keras.io/using-pre-trained-word-embeddings-in-a-keras-model.html.\n",
    "\n",
    "Installing the NLTK corpora can be a little tricky for the uninitiated. After installing the library, you must use the downloader to download one or more text corpora. The downloader is part of the library. This utility has a GUI which tends to run fine on Windows and Linux. On Mac, it is a better bet to use the command-line version by calling download_shell(), e.g.\n",
    "\n",
    "```bash\n",
    "$ python\n",
    "Python 3.7.6 (default, Jan  8 2020, 13:42:34)\n",
    "[Clang 4.0.1 (tags/RELEASE_401/final)] :: Anaconda, Inc. on darwin\n",
    "Type \"help\", \"copyright\", \"credits\" or \"license\" for more information.\n",
    ">>> import nltk\n",
    ">>> nltk.download_shell()\n",
    "```\n",
    "\n",
    "For this exercise, it probably makes sense to download the 'popular' collection of corpora ('d popular'). At the very least, you'll need the stop words ('d stopwords'). You'll also need the [GloVe pre-trained word vectors](http://nlp.stanford.edu/data/glove.6B.zip) from Stanford (careful, it's and 822MB download!).\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "from sklearn.datasets import fetch_20newsgroups\n",
    "from tensorflow.keras.preprocessing.text import Tokenizer\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras import layers\n",
    "from keras.utils.np_utils import to_categorical\n",
    "from nltk.corpus import stopwords\n",
    "from nltk.tokenize import word_tokenize\n",
    "from nltk.stem import PorterStemmer\n",
    "from nltk.tokenize import RegexpTokenizer\n",
    "import nltk\n",
    "import re"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Constants\n",
    "MAX_NB_WORDS = 20000 ## most commonly occuring words in the dataset\n",
    "MAX_SEQUENCE_LENGTH = 1000\n",
    "#VALIDATION_SPLIT = 0.2\n",
    "EMBEDDING_DIM = 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Read in some data\n",
    "categories = ['soc.religion.christian', 'comp.graphics', 'sci.med', 'comp.windows.x', 'sci.space', 'alt.atheism']\n",
    "\n",
    "#categories = ['alt.atheism',\n",
    "# 'comp.graphics',\n",
    "# 'comp.os.ms-windows.misc',\n",
    "# 'comp.sys.ibm.pc.hardware',\n",
    "# 'comp.sys.mac.hardware',\n",
    "# 'comp.windows.x',\n",
    "# 'misc.forsale',\n",
    "# 'rec.autos',\n",
    "# 'rec.motorcycles',\n",
    "# 'rec.sport.baseball',\n",
    "# 'rec.sport.hockey',\n",
    "# 'sci.crypt',\n",
    "# 'sci.electronics',\n",
    "# 'sci.med',\n",
    "# 'sci.space',\n",
    "# 'soc.religion.christian',\n",
    "# 'talk.politics.guns',\n",
    "# 'talk.politics.mideast',\n",
    "# 'talk.politics.misc',\n",
    "# 'talk.religion.misc']\n",
    "\n",
    "newsgroup_train = fetch_20newsgroups(subset='train', categories=categories, shuffle=True, remove=('headers', 'footers', 'quotes'))\n",
    "newsgroup_test = fetch_20newsgroups(subset='test', categories=categories, remove=('headers', 'footers', 'quotes'))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3443 documents\n",
      "6 categories\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(\"%d documents\" % len(newsgroup_train.filenames))\n",
    "print(\"%d categories\" % len(newsgroup_train.target_names))\n",
    "print()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "We have recently obtained a copy of color_xterm (from export.lcs.mit.edu) after\n",
      "seeing it mentioned in a previous article. On compilation, it reports the\n",
      "following undefined symbols:\n",
      "\t_get_wmShellWidgetClass\n",
      "\t_get_applicationShellWidgetClass\n",
      "\n",
      "\tbut still runs. When sending escape sequences to set the colour,\n",
      "any colour comes out as black text on a black background. Any Ideas?\n",
      "\n",
      "Has anyone used this program in colour successfully?\n",
      "\n",
      "Can anyone point me to any other colour terminal emulators?\n",
      "\n",
      "We are running OpenWindows 3 on Sun Sparcs running SunOS 4.1.3.\n",
      "\n",
      "\t\t\t\tadvaTHANKSnce  Ben Kelley.\n"
     ]
    }
   ],
   "source": [
    "# Print an example message\n",
    "print(newsgroup_train[\"data\"][1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Optional: Pre-Processing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\"mightn't\", 'very', \"wouldn't\", 'below', 'out', 'then', 'after', 'all', 'this', 'over', 'y', \"shan't\", \"won't\", 'haven', 'we', \"don't\", 'just', 'him', 'her', \"you'll\", 'as', 'herself', 'between', 'which', 'shan', \"hasn't\", 'wouldn', 'll', 'for', 'won', 've', 'these', 'both', 'me', 'm', 't', 'its', 'any', 'other', \"haven't\", 'hasn', 'did', 'does', 'than', 'you', 'from', 'there', 'be', 'wasn', \"hadn't\", 'ours', 'the', 'didn', \"you'd\", 'd', \"couldn't\", 'myself', 'theirs', 'such', 'do', 'had', 'their', 'no', \"wasn't\", \"isn't\", 'about', 'up', \"mustn't\", 'not', 'an', 'with', 'to', 'while', 'before', 'shouldn', 'has', 'yours', 'ma', 'how', 'most', 'itself', 'doesn', 'needn', 's', 'they', 'should', 'again', 'same', 'where', \"weren't\", 'whom', 'against', 'those', 'aren', 'through', 'down', 'my', 'off', \"you're\", 'but', 'weren', 'himself', 'further', \"shouldn't\", \"didn't\", 'and', 'is', 'yourself', 'mightn', \"you've\", 'your', 'that', 'now', 'mustn', 'them', 'why', 'were', \"it's\", 'was', \"needn't\", 'having', 'have', 'isn', 'once', 're', 'he', 'it', 'only', 'too', 'themselves', 'yourselves', 'at', 'own', \"doesn't\", 'in', 'few', 'are', \"she's\", 'on', 'each', 'here', 'who', 'am', 'into', \"aren't\", 'or', 'couldn', 'by', 'some', 'o', 'ain', 'doing', 'our', 'his', 'being', 'nor', 'i', 'so', 'if', 'more', \"should've\", 'because', \"that'll\", 'when', 'what', 'can', 'of', 'a', 'during', 'above', 'hadn', 'hers', 'under', 'ourselves', 'until', 'she', 'don', 'will', 'been'}\n"
     ]
    }
   ],
   "source": [
    "# Set up standard stop words\n",
    "stop_words = set(stopwords.words('english')) \n",
    "print(stop_words)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Put some additional words in the list\n",
    "stop_words.update(['find','would','be','one','say','made','think','subject','see','fnal','articl','feel','given','led',\n",
    "                  'view','way','either','however','yet','know','get','use','therefore','even','like','could'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " recently obtained copy color xterm export lcs mit edu seeing mentioned previous article compilation reports following undefined symbols get wmshellwidgetclass get applicationshellwidgetclass still runs sending escape sequences set colour colour comes black text black background ideas anyone used program colour successfully anyone point colour terminal emulators running openwindows sun sparcs running sunos advathanksnce ben kelley\n"
     ]
    }
   ],
   "source": [
    "# Pre-processing\n",
    "\n",
    "# Get rid of stop words e.g. with nltk (and stem)\n",
    "ps = PorterStemmer()\n",
    "tokenizer = RegexpTokenizer(r'\\w+') ## Std pattern is '\\w+|\\$[\\d\\.]+|\\S+'\n",
    "\n",
    "newsgroup_trainN = []\n",
    "\n",
    "\n",
    "for message in newsgroup_train.data:\n",
    "    \n",
    "    #r'[^A-Za-z0-9#]+'\n",
    "    \n",
    "    # Sub all the special characters (except _)\n",
    "    document = re.sub(r'\\W', ' ', message)\n",
    "    document = re.sub(r'_', ' ', document)\n",
    "    # Sub all single characters\n",
    "    document = re.sub(r'\\s+[a-zA-Z]\\s+', ' ', document)\n",
    "    # Sub single characters from the start\n",
    "    document = re.sub(r'\\^[a-zA-Z]\\s+', ' ', document) \n",
    "    # Sub digits (https://stackoverflow.com/questions/39582859/python-regex-remove-numbers-and-numbers-with-punctaution?rq=1)\n",
    "    document = re.sub(r'\\b\\d+(?:\\.\\d+)?\\s+', ' ', document)\n",
    "    \n",
    "    # Substituting multiple spaces with single space\n",
    "    document = re.sub(r'\\s+', ' ', document, flags=re.I)\n",
    "    # Convert to Lowercase\n",
    "    document = document.lower()\n",
    "    document = document.strip()\n",
    "    \n",
    "    # Original code\n",
    "    text_token = tokenizer.tokenize(document)\n",
    "    messageN = ''\n",
    "    \n",
    "    for word in text_token:\n",
    "        if word not in stop_words:\n",
    "            #messageN = messageN + ' ' + ps.stem(word)\n",
    "            # w/o stemming\n",
    "            messageN = messageN + ' ' + word\n",
    "            \n",
    "    newsgroup_trainN.append(messageN)\n",
    "\n",
    "newsgroup_train.data = newsgroup_trainN\n",
    "\n",
    "# Print the same message as above, now in its processed form\n",
    "print(newsgroup_train[\"data\"][1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Tokenization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 35942 unique tokens.\n",
      "[11738, 292, 330, 1073, 179, 2857, 57, 3085, 9875, 2161]\n"
     ]
    }
   ],
   "source": [
    "# Tokenize\n",
    "tokenizer = Tokenizer(num_words=MAX_NB_WORDS)\n",
    "tokenizer.fit_on_texts(newsgroup_train.data)\n",
    "\n",
    "# Get size of vocabulary \n",
    "# 1 needs to be added because of resrver 0 index\n",
    "vocab_size = len(tokenizer.word_index) + 1\n",
    "print('Found %s unique tokens.' % vocab_size)\n",
    "\n",
    "# Transforms each text in texts to a sequence of integers\n",
    "X_train = tokenizer.texts_to_sequences(newsgroup_train.data)\n",
    "X_test = tokenizer.texts_to_sequences(newsgroup_test.data)\n",
    "\n",
    "print (X_train[0][:10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of data tensor: (3443, 1000)\n",
      "Shape of label tensor: (3443, 6)\n",
      "[  662   704   300  1172   364  1624   364  1596   161  9879  4812  5465\n",
      "  4813   440   625    33 11745   952  6958    33  1929  2649   311    64\n",
      "    57  3747   394    59    74     2    33    14   163  1735     1 11746\n",
      "   833   220    95  1249  5842     3   222 14430  3204   403   361  6350\n",
      "  7680   214]\n"
     ]
    }
   ],
   "source": [
    "X_train = pad_sequences(X_train, padding='post', maxlen=MAX_SEQUENCE_LENGTH)\n",
    "X_test = pad_sequences(X_test, padding='post', maxlen=MAX_SEQUENCE_LENGTH)\n",
    "\n",
    "labels = to_categorical(np.asarray(newsgroup_train.target))\n",
    "print('Shape of data tensor:', X_train.shape)\n",
    "print('Shape of label tensor:', labels.shape)\n",
    "\n",
    "print (X_train[0][200:250])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# We don't need the full GloVe vocabulary, only the words we really need\n",
    "# Helper function for creating the embedding matrix\n",
    "\n",
    "def create_embedding_matrix(filepath, word_index, embedding_dim):\n",
    "    embedding_matrix = np.zeros((vocab_size, embedding_dim))\n",
    "\n",
    "    with open(filepath) as f:\n",
    "        for line in f:\n",
    "            word, *vector = line.split()\n",
    "            if word in word_index:\n",
    "                idx = word_index[word] \n",
    "                embedding_matrix[idx] = np.array(\n",
    "                    vector, dtype=np.float32)[:embedding_dim]\n",
    "\n",
    "    return embedding_matrix\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create the embedding matrix for our vocab\n",
    "#num_words = min(MAX_NB_WORDS, vocab_size)\n",
    "embedding_matrix = create_embedding_matrix('glove.6B.100d.txt', tokenizer.word_index, EMBEDDING_DIM)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.7691558622224696"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Check how many of our words are covered by the GloVe vocab\n",
    "nonzero_elements = np.count_nonzero(np.count_nonzero(embedding_matrix, axis=1))\n",
    "nonzero_elements / vocab_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_1 (Embedding)      (None, 1000, 100)         10537300  \n",
      "_________________________________________________________________\n",
      "conv1d_3 (Conv1D)            (None, 996, 128)          64128     \n",
      "_________________________________________________________________\n",
      "max_pooling1d_2 (MaxPooling1 (None, 199, 128)          0         \n",
      "_________________________________________________________________\n",
      "conv1d_4 (Conv1D)            (None, 195, 128)          82048     \n",
      "_________________________________________________________________\n",
      "max_pooling1d_3 (MaxPooling1 (None, 39, 128)           0         \n",
      "_________________________________________________________________\n",
      "conv1d_5 (Conv1D)            (None, 35, 128)           82048     \n",
      "_________________________________________________________________\n",
      "global_max_pooling1d_1 (Glob (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 128)               16512     \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 20)                2580      \n",
      "=================================================================\n",
      "Total params: 10,784,616\n",
      "Trainable params: 247,316\n",
      "Non-trainable params: 10,537,300\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "\n",
    "model.add(layers.Embedding(input_dim=vocab_size, output_dim=EMBEDDING_DIM, \n",
    "                           weights=[embedding_matrix], \n",
    "                           input_length=MAX_SEQUENCE_LENGTH, \n",
    "                           trainable=False))\n",
    "\n",
    "## --> 74% accuracy w/ GloVe\n",
    "\n",
    "model.add(layers.Conv1D(128, 5, activation='relu'))\n",
    "model.add(layers.MaxPooling1D(5))\n",
    "model.add(layers.Conv1D(128, 5, activation='relu'))\n",
    "model.add(layers.MaxPooling1D(5))\n",
    "model.add(layers.Conv1D(128, 5, activation='relu'))\n",
    "\n",
    "model.add(layers.GlobalMaxPool1D())\n",
    "\n",
    "model.add(layers.Dense(128, activation='relu'))\n",
    "model.add(layers.Dense(6, activation='softmax'))\n",
    "\n",
    "model.compile(optimizer='adam',\n",
    "              loss='sparse_categorical_crossentropy',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "model.summary()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 11314 samples, validate on 7532 samples\n",
      "Epoch 1/10\n",
      "11314/11314 [==============================] - 80s 7ms/sample - loss: 2.1313 - accuracy: 0.2758 - val_loss: 1.6949 - val_accuracy: 0.4222\n",
      "Epoch 2/10\n",
      "11314/11314 [==============================] - 79s 7ms/sample - loss: 1.3969 - accuracy: 0.5283 - val_loss: 1.4438 - val_accuracy: 0.5420\n",
      "Epoch 3/10\n",
      "11314/11314 [==============================] - 80s 7ms/sample - loss: 1.1096 - accuracy: 0.6254 - val_loss: 1.4434 - val_accuracy: 0.5607\n",
      "Epoch 4/10\n",
      "11314/11314 [==============================] - 92s 8ms/sample - loss: 0.9159 - accuracy: 0.6888 - val_loss: 1.5577 - val_accuracy: 0.5459\n",
      "Epoch 5/10\n",
      "11314/11314 [==============================] - 99s 9ms/sample - loss: 0.7509 - accuracy: 0.7405 - val_loss: 1.5952 - val_accuracy: 0.5720\n",
      "Epoch 6/10\n",
      "11314/11314 [==============================] - 94s 8ms/sample - loss: 0.6019 - accuracy: 0.7932 - val_loss: 1.7782 - val_accuracy: 0.5564\n",
      "Epoch 7/10\n",
      "11314/11314 [==============================] - 85s 7ms/sample - loss: 0.5117 - accuracy: 0.8259 - val_loss: 2.0729 - val_accuracy: 0.5501\n",
      "Epoch 8/10\n",
      "11314/11314 [==============================] - 86s 8ms/sample - loss: 0.4362 - accuracy: 0.8559 - val_loss: 2.2743 - val_accuracy: 0.5538\n",
      "Epoch 9/10\n",
      "11314/11314 [==============================] - 85s 8ms/sample - loss: 0.3706 - accuracy: 0.8748 - val_loss: 2.8073 - val_accuracy: 0.5223\n",
      "Epoch 10/10\n",
      "11314/11314 [==============================] - 86s 8ms/sample - loss: 0.3173 - accuracy: 0.8987 - val_loss: 2.5885 - val_accuracy: 0.5559\n",
      "Training Accuracy: 0.9191\n",
      "Testing Accuracy:  0.5559\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(X_train, newsgroup_train.target,\n",
    "                    epochs=10,\n",
    "                    verbose=True,\n",
    "                    validation_data=(X_test, newsgroup_test.target),\n",
    "                    batch_size=10)\n",
    "loss, accuracy = model.evaluate(X_train, newsgroup_train.target, verbose=False)\n",
    "print(\"Training Accuracy: {:.4f}\".format(accuracy))\n",
    "loss, accuracy = model.evaluate(X_test, newsgroup_test.target, verbose=False)\n",
    "print(\"Testing Accuracy:  {:.4f}\".format(accuracy))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEWCAYAAABrDZDcAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjMsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+AADFEAAAgAElEQVR4nO3deXxU9bnH8c+ThWwkISQBAiELiAKyE0FQRAWtW8VqRVGrooh4q7Va761dbldva/tqrRsVqeK+a22tda+KuxAU2beEBAIEspCFhKzz3D/OBIaQwACZnEnmeb9e88qcZWaeGfH3Ped3zvkdUVWMMcaErjC3CzDGGOMuCwJjjAlxFgTGGBPiLAiMMSbEWRAYY0yIsyAwxpgQZ0FgQoKIZImIikiEH+teKyKfdEZdxgQDCwITdESkQEQaRCSl1fzl3sY8y53KjOmeLAhMsNoMzGqZEJGRQIx75QQHf/ZojDlSFgQmWD0FXO0zfQ3wpO8KIpIoIk+KSImIFIrIz0UkzLssXET+JCKlIpIPnN/Gax8VkR0isk1E7hKRcH8KE5GXRKRYRCpF5CMROdFnWYyI/NlbT6WIfCIiMd5lp4rIZyJSISJbReRa7/wPRWSOz3sc0DXl3Qv6vohsBDZ6593nfY8qEVkmIlN81g8XkZ+KSJ6IVHuXDxSR+SLy51bf5V8i8kN/vrfpviwITLD6AkgQkWHeBvoy4OlW6zwAJAKDgKk4wTHbu+wG4AJgLJADfLfVa58AmoDjvOucDczBP28CQ4A+wFfAMz7L/gSMByYDvYH/ATwikuF93QNAKjAGWO7n5wFcBEwEhnunl3rfozfwLPCSiER7l92Oszd1HpAAXAfUer/zLJ+wTAGmAc8dQR2mO1JVe9gjqB5AATAd+Dnwe+Ac4F0gAlAgCwgH6oHhPq+7EfjQ+/x9YJ7PsrO9r40A+npfG+OzfBbwgff5tcAnftbay/u+iTgbVnuB0W2s9xPg1Xbe40Ngjs/0AZ/vff8zD1PH7pbPBdYDM9pZby1wlvf5zcAbbv/3tof7D+tvNMHsKeAjIJtW3UJACtADKPSZVwgM8D7vD2xttaxFJhAJ7BCRlnlhrdZvk3fv5P+AS3G27D0+9UQB0UBeGy8d2M58fx1Qm4j8CGcPpj9OUCR4azjcZz0BXIUTrFcB9x1DTaabsK4hE7RUtRDnoPF5wN9bLS4FGnEa9RYZwDbv8x04DaLvshZbcfYIUlS1l/eRoKoncnhXADNw9lgScfZOAMRbUx0wuI3XbW1nPkANEOsz3a+NdfYNE+w9HvBjYCaQpKq9gEpvDYf7rKeBGSIyGhgG/KOd9UwIsSAwwe56nG6RGt+ZqtoMvAj8n4jEi0gmTt94y3GEF4EfiEi6iCQBd/q8dgfwDvBnEUkQkTARGSwiU/2oJx4nRMpwGu/f+byvB1gE3CMi/b0HbSeJSBTOcYTpIjJTRCJEJFlExnhfuhy4WERiReQ473c+XA1NQAkQISK/wNkjaPEI8FsRGSKOUSKS7K2xCOf4wlPAK6q614/vbLo5CwIT1FQ1T1Vz21l8C87WdD7wCc5B00XeZX8D3ga+wTmg23qP4mqcrqU1OP3rLwNpfpT0JE430zbva79otfwOYCVOY1sO/AEIU9UtOHs2P/LOXw6M9r7mL0ADsBOn6+YZDu1tnAPPG7y11HFg19E9OEH4DlAFPMqBp94+AYzECQNjEFW7MY0xoURETsPZc8ry7sWYEGd7BMaEEBGJBG4FHrEQMC0sCIwJESIyDKjA6QK71+VyTBAJWBCIyCIR2SUiq9pZLiJyv4hsEpEVIjIuULUYY0BV16pqnKpOVtUqt+sxwSOQewSP41wI1J5zca7OHALMBR4KYC3GGGPaEbALylT1o8OMEjkDeFKdo9VfiEgvEUnzntrXrpSUFM3KOtTbGmOMaW3ZsmWlqpra1jI3rywewIGnvBV55x0UBCIyF2evgYyMDHJz2zub0BhjTFtEpLC9ZW4eLJY25rV5LquqLlTVHFXNSU1tM9CMMcYcJTeDoIgDhwBIB7a7VIsxxoQsN4PgNeBq79lDJwOVhzs+YIwxpuMF7BiBiDwHnA6kiEgR8EucER9R1QXAGziX3G/CGSt9dtvvdHiNjY0UFRVRV1d3rGV3GdHR0aSnpxMZGel2KcaYLi6QZw3NOsxyBb7fEZ9VVFREfHw8WVlZ+Awr3G2pKmVlZRQVFZGdne12OcaYLq5bXFlcV1dHcnJySIQAgIiQnJwcUntAxpjA6RZBAIRMCLQIte9rjAkcu0OZMcYEKVWlaPde1uyoYt2OaqYN68OIAYkd/jkWBB2grKyMadOmAVBcXEx4eDgt1zssWbKEHj16HPY9Zs+ezZ133skJJ5wQ0FqNMcGptqGJdcXVrNtRzdodVazdUcW64mr21DcBIALJPXtYEASr5ORkli9fDsCvfvUrevbsyR133HHAOi03iQ4La7s37rHHHgt4ncYY97Vs5a8rPrDBLyiroeX2MPFREQxNi+c7YwcwLC2BYWnxHN83nriowDTZFgQBtGnTJi666CJOPfVUvvzyS15//XV+/etf89VXX7F3714uu+wyfvGLXwBw6qmn8uCDDzJixAhSUlKYN28eb775JrGxsfzzn/+kT58+Ln8bY8yRqm1oYsPOPfsb/B3VrC2uorquad86mcmxDOuXwEVjBjA0LZ7haQmkJ8V06nHAbhcEv/7XatZs79gRdof3T+CX3/bnvuYHW7NmDY899hgLFiwA4O6776Z37940NTVxxhln8N3vfpfhw4cf8JrKykqmTp3K3Xffze23386iRYu4884723p7Y0wQUFW2V9axdvv+Lfy1O6rY7LOVH9cjnKFpCcwY05+h/RIYlpbA0H6B28o/Eu5X0M0NHjyYk046ad/0c889x6OPPkpTUxPbt29nzZo1BwVBTEwM5557LgDjx4/n448/7tSajTHtq2tsZr23oV9XXO09kFtFlc9WfkbvWIalxfPt0f0Zlpawbys/LCw4z/brdkFwtFvugRIXF7fv+caNG7nvvvtYsmQJvXr14qqrrmrzWgDfg8vh4eE0NTUdtI4x5vBUlSaP0tSsNHo8NDUrTc0eGpq9zz0eGpsPXt7o8f5t9tDQrGwtr2WNt3unoLQGj3crP7ZHOEP7xXPBvgbf6cuPj+5aV/x3uyAIZlVVVcTHx5OQkMCOHTt4++23OeecQ927x5jQ0dDkYevuWvJLathcuofNpTWU1zTQ2Kw0tm64vY12Y7NnX0PvNO6eAxt1T5sDGh+Vgb1jGNovgQtG9Wd4WjxD+yWQ0Ts2aLfyj4QFQScaN24cw4cPZ8SIEQwaNIhTTjnF7ZKM6VSqSnFVHZtLasgvrTmg0d+6ey/NPg1377gepPaMIiJciAgPIzJMiAwPIzrS+RvhnY4IFyLCwogMlwOeO8uc10WEe5cf8Nx5bWR42L73idw378Dl/RKjSehiW/lHQlQ7LjE7Q05Ojra+Mc3atWsZNmyYSxW5J1S/twl+lbWN5Hsb+M2l+xv9gtIa9jY271svOjKM7JSeDEqJIzsljkGpzt/slDh6xR7++hvjPxFZpqo5bS2zPQJjzFGpa2xmS3kt+SV7yC+tYXNJzb6Gv6ymYd964WHCwKQYslPimDQomezUOAZ5G/2+8dHdomulq7MgMMa0q9mjbK/Yu3/LvqXRL61hW8VefDsUUuOjyE6J4+wT+3q36nuSnRJHRu9YekR0m2HNuiULAmMM4Fz8tHxrBcsKdrN6exX5pXsoKKulocmzb52eURFkp8QxLiOJS8alMyg1jkEpPclKie1yZ8qY/SwIjAlROyr3kluwm2WFzmPNjqp9B2udrpuenH5Cn3199oNS40jtGWUj33ZDFgTGhICmZg/riqvJLShn2ZYKlhWUs73SuYYlJjKc0QMTuWnqYMZnJTFuYBKJsbZ1H0osCIzphir3NvL1lv1b+8u3VlDb4Jyt0y8hmvFZSdyQmcT4zCSGpSUQGW59+KHMgqADdMQw1ACLFi3ivPPOo1+/fgGr1XQ/qsqW8lqnm2fLbpYV7GbDrmpUIUxgWFoCl45PZ1xmEjlZvRnQK8btkk2QsSDoAP4MQ+2PRYsWMW7cOAsCc0j1Tc2s2lbFssJy7xZ/BaV76gFn+OKxmUmcPyqN8ZlJjBnYKygGNTPBzf6FBNgTTzzB/PnzaWhoYPLkyTz44IN4PB5mz57N8uXLUVXmzp1L3759Wb58OZdddhkxMTFHtCdhureyPfVOg+/d2l+xrXLfmTwZvWM5bUiKd2s/iSF94gm38/LNEep+QfDmnVC8smPfs99IOPfuI37ZqlWrePXVV/nss8+IiIhg7ty5PP/88wwePJjS0lJWrnTqrKiooFevXjzwwAM8+OCDjBkzpmPrN12Gx6Pklewht3B///7m0hoAIsOFEQMSuWZSJuMzkxiXmUSf+GiXKzbdQfcLgiDy3nvvsXTpUnJynKu69+7dy8CBA/nWt77F+vXrufXWWznvvPM4++yzXa7UuG1XdR2PfryZF3K3UlHbCDhj7YzLSGJmzkByspIYOSCR6Mhwlys13VH3C4Kj2HIPFFXluuuu47e//e1By1asWMGbb77J/fffzyuvvMLChQtdqNC4bXvFXh5enMfzS7fS2Ozh3JFpTD0+lZzMJLJT4uycfdMpul8QBJHp06fz3e9+l1tvvZWUlBTKysqoqakhJiaG6OhoLr30UrKzs5k3bx4A8fHxVFdXu1y16QyFZTU89GEer3xVBMAl49KZN3UwWSlxh3mlMR3PgiCARo4cyS9/+UumT5+Ox+MhMjKSBQsWEB4ezvXXX4+qIiL84Q9/AGD27NnMmTPHDhZ3Yxt3VvPXD/P45/JtRISHccWEDOZOHWyndBpX2TDUXViofu+uaPX2SuZ/sIk3VxUTExnOVSdnMufUbPok2MFe0zlsGGpjXPLVlt08+P4m3l+3i/ioCG4+4zhmn5JN7zjb2zPBw4LAmA6mqnyRX86DH2zk001lJMVGcsfZx/O9SVkkxtgYPib4BDQIROQc4D4gHHhEVe9utTwTWASkAuXAVapadDSf1dLfHiq6WpdeKFBVFm8o4cH3N5FbuJvU+Ch+dt4wrpiYYVf3mqAWsH+dIhIOzAfOAoqApSLymqqu8VntT8CTqvqEiJwJ/B743pF+VnR0NGVlZSQnJ4dEGKgqZWVlREdb/3Iw8HiUd9fu5MH3N7FyWyX9E6P5zYwTmZkz0M77N11CIDdTJgCbVDUfQESeB2YAvkEwHLjN+/wD4B9H80Hp6ekUFRVRUlJyDOV2LdHR0aSnp7tdRkhr9iivr9jO/A82sWHnHjKTY/njJaO4aOwAuyOX6VICGQQDgK0+00XAxFbrfANcgtN99B0gXkSSVbXMdyURmQvMBcjIyDjogyIjI8nOzu64yo05hMZmD69+vY2HPsxjc2kNQ/r05L7Lx3D+yDQibDhn0wUFMgja6qNp3bF9B/CgiFwLfARsA5oOepHqQmAhOKePdmyZxvinrrGZl5YVseDDPLZV7OXE/gksuGocZw/vZzdgN11aIIOgCBjoM50ObPddQVW3AxcDiEhP4BJVrQxgTcYcsdqGJp79cgsLP8pnV3U94zJ6cddFIzj9hNSQOCZlur9ABsFSYIiIZONs6V8OXOG7goikAOWq6gF+gnMGkTFBoaqukac+L+TRTzZTXtPApEHJ3HvZGCYNDo2TEkzoCFgQqGqTiNwMvI1z+ugiVV0tIr8BclX1NeB04PciojhdQ98PVD3G+Gt3TQOPfbqZxz4roLquidNPSOXmM44jJ6u326UZExDdYogJYzpCy1DQT31RSG1DM+ec2I/vn3EcI9MT3S7NmGNmQ0wYcwg7q+p46MM8nluyhcZmD98e3Z//Ov04TugX73ZpxnQKCwITskqq63nowzye/rIQj0e5eNwAbjr9OLJtKGgTYiwITMgpr2ng4Y/yePKzQhqaPVw8dgA/mDaEgb1j3S7NGFdYEJiQUVnbyN8+zuexTzdT29jMjNH9uXX68bYHYEKeBYHp9qrqGln0yWYe/Xgz1fVNnD8qjR9OG8KQvnYMwBiwIDDdWE19E49/VsDCj/Kp3NvI2cP7cttZxzMsLcHt0owJKhYEptvZ29DMU18UsGBxPuU1DZw5tA+3TT/eTgM1ph0WBKbbqGts5rklW/jrh3mUVNczZUgKt511POMyktwuzZigZkFguryGJg8v5G5l/vubKK6q4+RBvZl/xTgmZNuVwMb4w4LAdFmNzR7+/lUR9/9nE9sq9jI+M4l7Zo5m8nEpbpdmTJdiQWC6nGaP8o+vt3H/+xspLKtldHoiv7t4JKcNSbHB4Iw5ChYEpsvweJTXV+7g3vc2kF9Sw/C0BB65Oodpw/pYABhzDCwITNDzeJS3Vxfzl/c2sGHnHk7oG283hDGmA1kQmKClqvxn7S7ueXcDa3ZUMTg1jgdmjeX8kWkWAMZ0IAsCE3RUlcUbSvjLuxv4pqiSzORY7pk5mhljBhBuAWBMh7MgMEHls02l/PndDSwr3M2AXjH88ZJRfGfcACLtpvDGBIwFgQkKSzaXc8+76/kiv5x+CdHcddEIZuYMpEeEBYAxgWZBYFz1zdYK/vTOej7eWEpqfBS/+vZwLp+QQXRkuNulGRMyLAiMa179uog7XlpBr5hIfnbeMK46OZOYHhYAxnQ2CwLjiqe+KOR//7GKyYOTefh744mPjnS7JGNClgWB6XQPfZjHH95ax/RhfXnwirHWDWSMyywITKdRVf70znrmf5DHjDH9+dOlo+1sIGOCgAWB6RQej/Lrf63mic8LmTUhg7suGmHXBBgTJCwITMA1NXv48SsreeWrIm6Yks1PzxtmYwMZE0QsCExA1Tc188Pnl/PmqmJuP+t4bjnzOAsBY4KMBYEJmL0Nzdz49DI+2lDC/14wnOtPzXa7JGNMGywITEBU1TUy5/FclhaW84dLRnLZSRlul2SMaYcFgelw5TUNXLNoCWt3VHH/5WP59uj+bpdkjDkECwLToXZW1XHVI1+ypbyWhVeP58yhfd0uyRhzGBYEpsNsLa/lyke+pGxPPY/PnsCkwclul2SM8UNAr+YRkXNEZL2IbBKRO9tYniEiH4jI1yKyQkTOC2Q9JnA27drDpQs+p3JvI8/ccLKFgDFdSMCCQETCgfnAucBwYJaIDG+12s+BF1V1LHA58NdA1WMCZ9W2SmY+/DlNHuWFG09mzMBebpdkjDkCgdwjmABsUtV8VW0AngdmtFpHgQTv80RgewDrMQGQW1DOrL99QUxkOC/Nm8TQfgmHf5ExJqgE8hjBAGCrz3QRMLHVOr8C3hGRW4A4YHpbbyQic4G5ABkZdhpisPhkYyk3PJlLv8Ronp4zkQG9YtwuyRhzFAK5R9DW5aPaanoW8LiqpgPnAU+JyEE1qepCVc1R1ZzU1NQAlGqO1Duri7nu8aVkJsfy4o2TLASM6cICGQRFwECf6XQO7vq5HngRQFU/B6KBlADWZDrAP77exk3PfMXw/gm8MHcSqfFRbpdkjDkGgQyCpcAQEckWkR44B4Nfa7XOFmAagIgMwwmCkgDWZI7RM18WctuLy5mQ1Zun50wkMdZuKGNMVxewYwSq2iQiNwNvA+HAIlVdLSK/AXJV9TXgR8DfROQ2nG6ja1W1dfeRCRIPL87j92+uY9rQPsy/cpzdUMaYbiKgF5Sp6hvAG63m/cLn+RrglEDWYI6dqnLPuxt44P1NfHt0f+6ZaTeUMaY7sSuLzSF5PMpvXl/D458VcPlJA/m/74y0G8oY081YEJh2NXuUO19ZwUvLiphzajY/O99uKGNMd2RBYNrU0OThtheW8++VO/jh9CHcOm2IhYAx3ZQFgTnI3oZmbnpmGR+uL+Hn5w9jzpRBbpdkjAkgCwJzgOq6Rq5/IpelBeX8/uKRzJpgV3Ib090d9tQPEblZRJI6oxjjrt01DVz5yJd8Vbib+y4fayFgTIjw5xzAfsBSEXnRO6y0dRR3Q7uq6rhs4eesK67m4e+N50K7q5gxIeOwQaCqPweGAI8C1wIbReR3IjI4wLWZTrK1vJZLH/6cot17eXz2SUwbZncVMyaU+HVVkPdq32LvowlIAl4WkT8GsDbTCfJK9jDz4c/ZXdPA03MmMnmwDfVkTKg57MFiEfkBcA1QCjwC/LeqNnpHCd0I/E9gSzSBsnp7JVc/ugQReOHGSQxLs3sJGBOK/DlrKAW4WFULfWeqqkdELghMWSbQlhXu5trHlhAfFcHTcyYyKLWn2yUZY1ziT9fQG0B5y4SIxIvIRABVXRuowkzgbNpVzbWPLSE5rgcvzptkIWBMiPMnCB4C9vhM13jnmS6odE89sx9fSlREOE/PmUh6UqzbJRljXOZPEIjv0NCq6sEuROuS6hqbmftkLiXV9Tx6TY6FgDEG8K9Bz/ceMG7ZC/gvID9wJZlAUFX+++UVfLWlgoeuHMfogb3cLsmY4LZnFxR+Blu+gKY6iE6AqASITnQeUQnOPN/nPeIhrOsN0e5PEMwD7gd+jnPzmP/gvZG86Tr+8t5G/vXNdn58zlDOHZnmdjnGlypUbYfiFVCxFRLSoFcmJGU6jYzpHJXbnIa/8BPnb+kGZ35kLPSIg7pKaG44zJtI2wGxL0BaP088eH5kDHTydbuHDQJV3YVzm0nTRb36dRH3/2cjM3PSmTfVBpBzVXMTlG2C4pVQ/I3370qoLWt7/ehekJTlhEJLOPTyTicOhMjozqy++1CF3QXehv9T57G7wFkWlQAZJ8OYKyHrVEgbDeHeW7I21kF9FdRVQX2lEw51Vd557Tyv2g51a/e/TpsPXVtYRBth4g2M0bMge0qH/xz+XEcQjXOT+RNx7ikMgKpe1+HVmA63ZHM5P355JZMGJXPXRSNtKOnO1FALO1c7W/rFK5wGf+caaNrrLA/vAX2Gw9Dzod8o6DfSafSrd8DuQqdhqih0nu9cDevfbLVFKhCf1iokMvcHR3wahNntRAGn4S/duH9rv/AzqNrmLItJgsxTYMKNkDnZ+e/Q3u8WGe08evY5+joaanzCpCU0Kg8dJuWbnensqUf3uYfhT9fQU8A64FvAb4ArATtttAsoKK3hxqdySe8dw4KrxtMjouv1XXYZNaVOY7/D2+AXr3C2/NXjLI9OdBr7nOsgzdvopxy/f0vTV3w/6D/24PkeD+wpdgJid+H+kKgohIJPYMULOL23XmGR0Gvg/pBIyjpwryK2d6d3QXQajwd2rTmwq6emxFnWs6/T4Gee4jxSh3Zev74IRPV0HgnBM56XP0FwnKpeKiIzVPUJEXkW54b0JohV1jZy3eNLAVh0zUkkxrbR4Jgjpwq7N+/v0mlp+Ku3718ncaDT0J94sfM3bZQz71gb3bAwp/FI6O80ZK011UNl0YF7Ei1/17wGe8sPXL9HzwP3IHz3KhLTISq+6wRFc5MTvvu6ej6DugpnWUI6DD5zf8OfPLjrfK9O4k8QNHr/VojICJzxhrICVpE5Zg1NHuY9vYyi3Xt5es5EslLi3C6pa2pqgJJ1+7t1dqyAnaucXXUACYfUE5w+25aunX4jnS1tN0REOY1ccjvjQdZXH7wnsbsQyvMh/wNorD1w/fAeEJsCscnOd4preZ7iTMcmHzyvrT2cQGhqgO1f79/a3/IlNFQ7y3oPgmEXQOapTmAmZXZOTV2YP0Gw0Hs/gp8DrwE9gf8NaFXmqKkqP3t1JZ/nl/GXy0YzIdulRqmrqavav5Xf0qe/ax14vNtBkbHQdwSMmult8EdBn2HOGR5dRVQ89BvhPFpTdbq3KrzHJqq2OQewa8ugxvt3+3KoLXX6qtv9jESIS/YJh7ZCxGdeVIJ/W+eNe6Eod/+B3a1L9x9rSR0Koy71bvFPDqoul67ikEHgHViuSlV3Ax8BdspJkHtocR4vLSviB9OG8J2x6W6XE7zq9zjnh2/+EDZ/5Gztt/Svx6U6Df2kad7+/FHOVmZ3PvAqAj1TnUd6zqHXbW6E2vL9QVFb6v1b7oRJy7yqIidQa0qhub7t9wqL9AZDO4GxZ6fT8G9b5j1QLk6Qjb9mf8MfZyPmHqtDBoF3YLmbgRc7qR5zDN5YuYM/vrWeC0f357bpQ9wuJ7g01UPRUqfRz18M23LB0+R0f6RPgNPvhAHjna39+H5uVxvcwiMhvq/z8EfLmTLtBYbvvOJVzry9u53XSjj0HwMTb3S6ejImOmf5mA7lT9fQuyJyB/ACzjhDAKhqefsvMZ1t+dYKbnthOeMzk/jjd0fZaaLNTbDjG9i82Gn8t3zhdCVIGKSNgcm3QPZpMPBk6GFDbQSU75kySVn+vaa5yQmDyBjndSag/AmClusFvu8zT7FuoqBRtLuWOU/k0ichioXfG090ZDfuwmiPKuxau7/hL/jUueAHnHP1x1/rNPyZkyHGhtcIeuERTjeV6RT+XFmc3RmFmKNTXdfI9Y/nUt/UzPNzJ5LcM+rwL2pqgM/uh4otB16E1CvT6W/tCnsTLadx5rc0/B/vP088KRtOvAgGTYWsKUd/8Y8xIcKfK4uvbmu+qj7Z8eWYI9HU7OHmZ78mr2QPT1w3geP6xB/+RWV58Mr1zql3sckHD20QGdfGlao+f6P8+IxAqdoOmz/ev9VfudWZ37Ofc5549mnOo1eGezUa0wX50zV0ks/zaGAa8BVgQeAiVeXX/1rD4g0l/P7ikZxy3GHOnFCFb56Df9/hHOyb+RQMv9A5e6Ziy8Hnlu8ucLayG/Yc+D6xye2ERJZz0VREj477krXlTg0tB3jLNjrzY5KcLf1TbnUuuU8Z0jX2YowJUv50Dd3iOy0iiTjDThyWiJwD3AeEA4+o6t2tlv8FOMM7GQv0UVXrwPXD458V8NQXhdx42iBmTTjMFnBdJbx+O6x62Tnz4uKHnStHwTkQ13e482hN1WmMKwp8QsL7fMc3sPb1/efZAyCQMKD9sW969jv0pfz11VD4+f4t/uKVgDpXwGZOdk4ZzD4N+o7skkP9GhOsjuYGM7XAYc9NFJFwYD5wFlAELBWR11R1Tcs6qnqbz/q3AG0MsGJa+8/anfz29TWcPbwvPz5n6KFX3rrU6QqqLIIzfjWpXToAABOFSURBVA5Tbvf/fHgR5+KguGTn1MrWPM1tD5BWUQj5HzrLfMe+CY86cOyblqEMStY5Df+2ZftP6Rw4Ec74mdPwDxjXeVesGhOC/DlG8C/2/98cBgzHv+sKJgCbVDXf+z7PAzOANe2sPwv4pR/vG9JWb6/klue+5sT+idx7+RjCwtrpEvE0wyf3wAe/h8QBcN1bMHBCxxYTFu405InpkHXKwcsb65x+/N2FrfYqCp1Gv2UsGAmD/uNg8g+cA7wDJ3atK3aN6eL82SP4k8/zJqBQVYv8eN0AYKvPdBEwsa0VRSQTyAbe9+N9Q9bOqjqufzyXxJhIHrkmh9ge7fznq9wGr97o9K+PuAQu+Is7NziJjHb671Pa2YGsq3T2VBLT7QYsxrjInyDYAuxQ1ToAEYkRkSxVLTjM69raVNU25oFz45uXVdu+Y4OIzMV7V7SMjNA8I6S2oYnrn1hKdV0jL82bTN+Edm5IsvZ1eO1m5xTRGX+FMVcE74HUlpttGGNc5c8Rt5cAj890s3fe4RQBA32m04Ht7ax7OfBce2+kqgtVNUdVc1JTQ+8ik2aPcuvzy1mzvYoHrhjL8P4JB6/UUAuv3wYvXOn0vd/4EYy9MnhDwBgTNPzZI4hQ1X23RVLVBhHx5xzBpcAQEckGtuE09le0XklETgCSgM/9Kzn0/OGtdby7Zie//PZwzhzaxvguO1fDy9c5B10n3wJn/qJjT+M0xnRr/uwRlIjIhS0TIjIDKD3ci1S1CbgZ5yY2a4EXVXW1iPzG9/1wDhI/r6rtdRuFtGe/3MLCj/K5elIm107OOnChKny5EBae4ZzmedXf4ey7LASMMUdEDtf+ishg4BmgZZDvIuBqVd0U4NralJOTo7m5uW58dKf7ZGMp1zy2hClDUnjk6hwiwn1yu6YU/vl92PAWDDnbOR5gY7MYY9ohIstUtc0xxv25oCwPOFlEeuIER3VHF2gOtnFnNTc9s4whfXrywKyxB4ZA/ofw9xudWw+eczdMnGfHAowxR+2wXUMi8jsR6aWqe1S1WkSSROSuziguVJXuqWf240uJigjn0WtPIj7aezFVUwO8+0t48iKIToAb3oeTb7IQMMYcE3+OEZyrqhUtE967lZ0XuJJCW11jMzc8mUvpnnoevSaHAb28F1aV5cGis+HTe52hFuYudm6iYowxx8ifs4bCRSRKVevBuY4A8GOsY3OkPB7ljpe+4estFTx05ThGD/QOu/TN8/DvHzlX8s58EobPcLdQY0y34k8QPA38R0Qe807PBp4IXEmh6973NvD6ih38+JyhnDsyzbmh+r9/BCtfhIzJcPFCZ6weY4zpQP4cLP6jiKwApuNcLfwWkBnowkLNK8uKuP/9TczMSWfe1EGtBov7GUz5Ufe+eboxxjX+jj5ajHN18UxgM/BKwCoKQV/ml3Hn31cwaVAyd104HPn4z/DB75whnWe/6dyw2xhjAqTdIBCR43GuBp4FlOHcvF5U9Yz2XmOOXEFpDTc+vYyBvWN5eEZ/ejz7HWewuBMvdgaLs/vrGmMC7FB7BOuAj4Fvt1w8JiK3HWJ9c4Qqahu47vGlCPDclDISHr/WO1jcfBhj4wQZYzrHoYLgEpw9gg9E5C3gedoeUdQchYYmD/OeXkbJ7kreG/EOfd94GtJGwyWLIOU4t8szxoSQdoNAVV8FXhWROOAi4Dagr4g8BLyqqu90Uo3djqry01dXsnvzcj5NeYSE9Ztg0s0w7RcQYWfmGmM6lz9nDdXgjDX0jIj0Bi4F7gQsCI7SXz/YRMzyRfw7+lkiPL3gqlfguOlul2WMCVFHdM9iVS0HHvY+zFF4P3clx39wK9+PXIYOPgsuesgGizPGuOpobl5vjlRTA2x8B8/Xz3DahrfRcKHxrP8jctJ/QZg/o3wYY0zgWBAEiioUr4Dlz8LKl6C2jIaoZJ5oOofhF9zMlJMnu12hMcYAFgQdb88uWPGiEwC7VkN4DzjhPHTMFVzyRhR1PeDdCZPcrtIYY/axIOgITfXODWKWPwsb3wVthgHj4fw/OxeGxfbmow0lrC5ewh8vGUVYmJ2Fa4wJHhYER0sVtn/tNP6rXoa9u6FnP+eewWOugNQTDlh9wYd59E2IYsbY/u28oTHGuMOC4EhVF8OKF5wAKFkH4VEw7AIYfQUMOh3CD/5Jv9lawef5Zfz0vKFERdjAccaY4GJB4I/GOlj/htP45/0H1APpE+CCe+HE7xx2PKAFi/OIj45g1oSMTirYGGP8Z0HQHlXYtgyWPwOrXoG6Smc00FNvg9GzIGWIX2+TX7KHt1YXc9PUwftvOWmMMUHEgqC1qu3OHcG+eQ5KN0BENAy7EMbMguypR3xPgL99nE9keBizT8kOUMHGGHNsLAgAGvfCun87XT/5HzhdPxmT4Nv3w4kXQXTiUb3trqo6Xlm2je/mpJMab2MIGWOCU+gGgSoULfV2/bwK9ZWQONC5E9joWZA8+Jg/YtGnBTR5PMydMqgDCjbGmMAIvSCoLHK6fpY/C+V5EBnr3Ax+9CzImtJhQz5U1TXyzBeFnDsijayUuA55T2OMCYTQCYK8D+DTeyF/MaCQeQpMud0Jgaj4Dv+4Z7/cQnV9E/OmHvuehTHGBFLoBEFlEZTnw9Qfw+jLoXfgDt7WNzWz6JPNnHJcMiPTj+74gjHGdJbQCYLRlzu3f+yE0T5f/Wobu6rr+fPM0QH/LGOMOVahEwThnXMOf7NHWfhRPif2T+DU41I65TONMeZY2GD4HezdNcXkl9Ywb+pgxG4+b4zpAgIaBCJyjoisF5FNInJnO+vMFJE1IrJaRJ4NZD2Bpqo8tDifjN6xnDuin9vlGGOMXwLWNSQi4cB84CygCFgqIq+p6hqfdYYAPwFOUdXdItInUPV0hi/yy/lmawW/vWgEEeG2s2WM6RoC2VpNADapar6qNgDPAzNarXMDMF9VdwOo6q4A1hNwCxbnkdKzB5eOT3e7FGOM8Vsgg2AAsNVnusg7z9fxwPEi8qmIfCEi57T1RiIyV0RyRSS3pKQkQOUemzXbq1i8oYRrJ2cRHWlDTRtjuo5ABkFbR0q11XQEMAQ4HZgFPCIiB43prKoLVTVHVXNSU1M7vNCO8PBHecT1COd7J2e5XYoxxhyRQAZBETDQZzod2N7GOv9U1UZV3QysxwmGLmVreS2vr9jBrAkZJMbaUNPGmK4lkEGwFBgiItki0gO4HHit1Tr/AM4AEJEUnK6i/ADWFBCPfJxPmMD1U2yoaWNM1xOwIFDVJuBm4G1gLfCiqq4Wkd+IyIXe1d4GykRkDfAB8N+qWhaomgKhbE89L+RuZcaYAaQlxrhdjjHGHLGAXlmsqm8Ab7Sa9wuf5wrc7n10SU98Xkhdo4d5U22oaWNM12Qnux+D2oYmnvy8gOnD+nJcn44fwdQYYzqDBcExeH7JVipqG7npdNsbMMZ0XRYER6mx2cOjn2zmpKwkxmf2drscY4w5ahYER+n1FdvZVrHXbjxjjOnyLAiOgqry8OJ8ju/bkzNO6NLDIxljjAXB0fhwfQnriquZe9pgwsJsqGljTNdmQXAUHlqcR1piNBeO7u92KcYYc8wsCI7QV1t2s2RzOdefmk2PCPv5jDFdn7VkR2jBh3kkxkQya0KG26UYY0yHsCA4Apt27eHdtTu5elImcVGhc7tnY0z3ZkFwBBZ+lEeP8DCumZzldinGGNNhLAj8VFxZx6tfb2NmzkBSeka5XY4xxnQYCwI/Lfp0M80e5YYpNpyEMaZ7sSDwQ+XeRp79cgvnj+pPRnKs2+UYY0yHsiDww9NfFLKnvokbT7O9AWNM92NBcBh1jc089mkBU4akMGJAotvlGGNMh7MgOIxXviqidE89N9ngcsaYbsqC4BCaPcrfPspnVHoikwYnu12OMcYEhAXBIby1qpiCslrmTR2MiA0uZ4zpniwI2qGqLFicR3ZKHN86sZ/b5RhjTMBYELTjs7wyVm6r5IYpgwi3oaaNMd2YBUE7FizOI6VnFBePG+B2KcYYE1AWBG1Yta2SjzeWct2pWURHhrtdjjHGBJQFQRsWLM6jZ1QEV07MdLsUY4wJOAuCVgrLanhj5Q6unJhBYkyk2+UYY0zAWRC08reP84kIC+O6U7PdLsUYYzqFBYGP0j31vJRbxHfGDqBvQrTb5RhjTKewIPDx+KcFNDR7mDvVBpczxoQOCwKvPfVNPPl5AWcP78vg1J5ul2OMMZ3GgsDr+SVbqKprYp4NLmeMCTEBDQIROUdE1ovIJhG5s43l14pIiYgs9z7mBLKe9jQ0eXjk481MzO7N2IwkN0owxhjXRATqjUUkHJgPnAUUAUtF5DVVXdNq1RdU9eZA1eGPfy7fRnFVHb+/ZKSbZRhjjCsCuUcwAdikqvmq2gA8D8wI4OcdFY9HefijfIb2i+f041PdLscYYzpdIINgALDVZ7rIO6+1S0RkhYi8LCID23ojEZkrIrkikltSUtKhRb6/bhebdu2xoaaNMSErkEHQVquqrab/BWSp6ijgPeCJtt5IVReqao6q5qSmduxW+4LFeQzoFcMFo9I69H2NMaarCGQQFAG+W/jpwHbfFVS1TFXrvZN/A8YHsJ6D5BaUk1u4mxumZBMRbidQGWNCUyBbv6XAEBHJFpEewOXAa74riIjvZviFwNoA1nOQBYvzSIqNZOZJbfZIGWNMSAjYWUOq2iQiNwNvA+HAIlVdLSK/AXJV9TXgByJyIdAElAPXBqqe1jbsrOa9tbv44fQhxPYI2M9gjDFBL6AtoKq+AbzRat4vfJ7/BPhJIGtoz8OL84mJDOeaSVlufLwxxgSNkOwY316xl38u38ZlJw0kKa6H2+UYY4yrQjIIHv1kMwrMmWJDTRtjTMgFQUVtA88t2cKFo/uTnhTrdjnGGOO6kAuCpz4vpLahmRttqGljjAFCLAjqGpt5/LMCzjghlaH9EtwuxxhjgkJIBcFLuVspq2mwoaaNMcZHyARBU7OHhR/nMzajFxOye7tdjjHGBI2QCYI3VhWztXyvDS5njDGthEwQxPUI56zhfTlrWF+3SzHGmKASMmMrTBvWl2kWAsYYc5CQ2SMwxhjTNgsCY4wJcRYExhgT4iwIjDEmxFkQGGNMiLMgMMaYEGdBYIwxIc6CwBhjQpyoqts1HBERKQEKj/LlKUBpB5bT1dnvcSD7Pfaz3+JA3eH3yFTV1LYWdLkgOBYikquqOW7XESzs9ziQ/R772W9xoO7+e1jXkDHGhDgLAmOMCXGhFgQL3S4gyNjvcSD7Pfaz3+JA3fr3CKljBMYYYw4WansExhhjWrEgMMaYEBcyQSAi54jIehHZJCJ3ul2PW0RkoIh8ICJrRWS1iNzqdk3BQETCReRrEXnd7VrcJiK9RORlEVnn/Xcyye2a3CIit3n/P1klIs+JSLTbNQVCSASBiIQD84FzgeHALBEZ7m5VrmkCfqSqw4CTge+H8G/h61ZgrdtFBIn7gLdUdSgwmhD9XURkAPADIEdVRwDhwOXuVhUYIREEwARgk6rmq2oD8Dwww+WaXKGqO1T1K+/zapz/yQe4W5W7RCQdOB94xO1a3CYiCcBpwKMAqtqgqhXuVuWqCCBGRCKAWGC7y/UERKgEwQBgq890ESHe+AGISBYwFvjS3Upcdy/wP4DH7UKCwCCgBHjM21X2iIjEuV2UG1R1G/AnYAuwA6hU1XfcrSowQiUIpI15IX3erIj0BF4BfqiqVW7X4xYRuQDYparL3K4lSEQA44CHVHUsUAOE5DE1EUnC6TnIBvoDcSJylbtVBUaoBEERMNBnOp1uuovnDxGJxAmBZ1T1727X47JTgAtFpACny/BMEXna3ZJcVQQUqWrLXuLLOMEQiqYDm1W1RFUbgb8Dk12uKSBCJQiWAkNEJFtEeuAc8HnN5ZpcISKC0/+7VlXvcbset6nqT1Q1XVWzcP5dvK+q3XKrzx+qWgxsFZETvLOmAWtcLMlNW4CTRSTW+//NNLrpgfMItwvoDKraJCI3A2/jHPlfpKqrXS7LLacA3wNWishy77yfquobLtZkgsstwDPejaZ8YLbL9bhCVb8UkZeBr3DOtvuabjrUhA0xYYwxIS5UuoaMMca0w4LAGGNCnAWBMcaEOAsCY4wJcRYExhgT4iwIjGlFRJpFZLnPo8OurBWRLBFZ1VHvZ0xHCInrCIw5QntVdYzbRRjTWWyPwBg/iUiBiPxBRJZ4H8d552eKyH9EZIX3b4Z3fl8ReVVEvvE+WoYnCBeRv3nHuX9HRGJc+1LGYEFgTFtiWnUNXeazrEpVJwAP4oxaivf5k6o6CngGuN87/35gsaqOxhmvp+Vq9iHAfFU9EagALgnw9zHmkOzKYmNaEZE9qtqzjfkFwJmqmu8duK9YVZNFpBRIU9VG7/wdqpoiIiVAuqrW+7xHFvCuqg7xTv8YiFTVuwL/zYxpm+0RGHNktJ3n7a3Tlnqf583YsTrjMgsCY47MZT5/P/c+/4z9tzC8EvjE+/w/wE2w757ICZ1VpDFHwrZEjDlYjM/IrODcv7flFNIoEfkSZyNqlnfeD4BFIvLfOHf3ahmt81ZgoYhcj7PlfxPOna6MCSp2jMAYP3mPEeSoaqnbtRjTkaxryBhjQpztERhjTIizPQJjjAlxFgTGGBPiLAiMMSbEWRAYY0yIsyAwxpgQ9/+wkDlYEzDdowAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot training & validation accuracy values\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "# Plot training & validation loss values\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('Model loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Same model but w/o GloVe\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_4\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_3 (Embedding)      (None, 1000, 100)         4004900   \n",
      "_________________________________________________________________\n",
      "conv1d_9 (Conv1D)            (None, 996, 128)          64128     \n",
      "_________________________________________________________________\n",
      "max_pooling1d_6 (MaxPooling1 (None, 199, 128)          0         \n",
      "_________________________________________________________________\n",
      "conv1d_10 (Conv1D)           (None, 195, 128)          82048     \n",
      "_________________________________________________________________\n",
      "max_pooling1d_7 (MaxPooling1 (None, 39, 128)           0         \n",
      "_________________________________________________________________\n",
      "conv1d_11 (Conv1D)           (None, 35, 128)           82048     \n",
      "_________________________________________________________________\n",
      "global_max_pooling1d_3 (Glob (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 128)               16512     \n",
      "_________________________________________________________________\n",
      "dense_7 (Dense)              (None, 6)                 774       \n",
      "=================================================================\n",
      "Total params: 4,250,410\n",
      "Trainable params: 4,250,410\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "\n",
    "model.add(layers.Embedding(input_dim=vocab_size, output_dim=EMBEDDING_DIM, \n",
    "                           input_length=MAX_SEQUENCE_LENGTH))\n",
    "\n",
    "## --> 70% accuracy w/o GloVe\n",
    "\n",
    "\n",
    "model.add(layers.Conv1D(128, 5, activation='relu'))\n",
    "model.add(layers.MaxPooling1D(5))\n",
    "model.add(layers.Conv1D(128, 5, activation='relu'))\n",
    "model.add(layers.MaxPooling1D(5))\n",
    "model.add(layers.Conv1D(128, 5, activation='relu'))\n",
    "\n",
    "model.add(layers.GlobalMaxPool1D())\n",
    "\n",
    "model.add(layers.Dense(128, activation='relu'))\n",
    "model.add(layers.Dense(6, activation='softmax'))\n",
    "\n",
    "model.compile(optimizer='adam',\n",
    "              loss='sparse_categorical_crossentropy',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "model.summary()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 3443 samples, validate on 2291 samples\n",
      "Epoch 1/10\n",
      "3443/3443 [==============================] - 47s 14ms/sample - loss: 1.5025 - accuracy: 0.3218 - val_loss: 1.2391 - val_accuracy: 0.4391\n",
      "Epoch 2/10\n",
      "3443/3443 [==============================] - 46s 13ms/sample - loss: 0.9339 - accuracy: 0.5937 - val_loss: 1.0534 - val_accuracy: 0.5740\n",
      "Epoch 3/10\n",
      "3443/3443 [==============================] - 46s 13ms/sample - loss: 0.5565 - accuracy: 0.7766 - val_loss: 1.1207 - val_accuracy: 0.6264\n",
      "Epoch 4/10\n",
      "3443/3443 [==============================] - 46s 13ms/sample - loss: 0.3427 - accuracy: 0.8687 - val_loss: 1.2797 - val_accuracy: 0.6552\n",
      "Epoch 5/10\n",
      "3443/3443 [==============================] - 47s 14ms/sample - loss: 0.1945 - accuracy: 0.9306 - val_loss: 1.3693 - val_accuracy: 0.6792\n",
      "Epoch 6/10\n",
      "3443/3443 [==============================] - 47s 14ms/sample - loss: 0.0923 - accuracy: 0.9646 - val_loss: 1.7220 - val_accuracy: 0.6779\n",
      "Epoch 7/10\n",
      "3443/3443 [==============================] - 47s 14ms/sample - loss: 0.0819 - accuracy: 0.9689 - val_loss: 1.7647 - val_accuracy: 0.6940\n",
      "Epoch 8/10\n",
      "3443/3443 [==============================] - 47s 14ms/sample - loss: 0.0937 - accuracy: 0.9646 - val_loss: 1.6371 - val_accuracy: 0.7080\n",
      "Epoch 9/10\n",
      "3443/3443 [==============================] - 47s 14ms/sample - loss: 0.0929 - accuracy: 0.9663 - val_loss: 1.5760 - val_accuracy: 0.7014\n",
      "Epoch 10/10\n",
      "3443/3443 [==============================] - 46s 13ms/sample - loss: 0.0821 - accuracy: 0.9689 - val_loss: 1.6950 - val_accuracy: 0.7041\n",
      "Training Accuracy: 0.9791\n",
      "Testing Accuracy:  0.7041\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(X_train, newsgroup_train.target,\n",
    "                    epochs=10,\n",
    "                    verbose=True,\n",
    "                    validation_data=(X_test, newsgroup_test.target),\n",
    "                    batch_size=10)\n",
    "loss, accuracy = model.evaluate(X_train, newsgroup_train.target, verbose=False)\n",
    "print(\"Training Accuracy: {:.4f}\".format(accuracy))\n",
    "loss, accuracy = model.evaluate(X_test, newsgroup_test.target, verbose=False)\n",
    "print(\"Testing Accuracy:  {:.4f}\".format(accuracy))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot training & validation accuracy values\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "# Plot training & validation loss values\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('Model loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Back to Michaels best model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_2 (Embedding)      (None, 1000, 100)         3594200   \n",
      "_________________________________________________________________\n",
      "global_max_pooling1d_2 (Glob (None, 100)               0         \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 10)                1010      \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 10)                0         \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 6)                 66        \n",
      "=================================================================\n",
      "Total params: 3,595,276\n",
      "Trainable params: 1,076\n",
      "Non-trainable params: 3,594,200\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# dropout\n",
    "\n",
    "# another way to work with embeddings is by using a MaxPooling1D/AveragePooling1D\n",
    "# or a GlobalMaxPooling1D/GlobalAveragePooling1D layer after the embedding.\n",
    "# You can think of the pooling layers as a way to downsample (a way to reduce the size of)\n",
    "# the incoming feature vectors\n",
    "\n",
    "#embedding_dim = 50\n",
    "\n",
    "model = Sequential()\n",
    "model.add(layers.Embedding(input_dim=vocab_size, output_dim=EMBEDDING_DIM, \n",
    "                           weights=[embedding_matrix], \n",
    "                           input_length=MAX_SEQUENCE_LENGTH, \n",
    "                           trainable=False))\n",
    "\n",
    "model.add(layers.GlobalMaxPool1D())\n",
    "model.add(layers.Dense(10, activation='relu'))\n",
    "model.add(layers.Dropout(0.3))\n",
    "model.add(layers.Dense(6, activation='softmax'))\n",
    "model.compile(optimizer='adam',\n",
    "              loss='sparse_categorical_crossentropy',\n",
    "              metrics=['accuracy'])\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 3443 samples, validate on 2291 samples\n",
      "Epoch 1/50\n",
      "3443/3443 [==============================] - 2s 579us/sample - loss: 1.7679 - accuracy: 0.2001 - val_loss: 1.7209 - val_accuracy: 0.2807\n",
      "Epoch 2/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.6950 - accuracy: 0.2881 - val_loss: 1.6429 - val_accuracy: 0.3470\n",
      "Epoch 3/50\n",
      "3443/3443 [==============================] - 2s 438us/sample - loss: 1.6152 - accuracy: 0.3294 - val_loss: 1.5616 - val_accuracy: 0.3715\n",
      "Epoch 4/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.5487 - accuracy: 0.3477 - val_loss: 1.4980 - val_accuracy: 0.4116\n",
      "Epoch 5/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.4938 - accuracy: 0.3674 - val_loss: 1.4405 - val_accuracy: 0.4208\n",
      "Epoch 6/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.4133 - accuracy: 0.4069 - val_loss: 1.3547 - val_accuracy: 0.4579\n",
      "Epoch 7/50\n",
      "3443/3443 [==============================] - 2s 438us/sample - loss: 1.3908 - accuracy: 0.4014 - val_loss: 1.3012 - val_accuracy: 0.4684\n",
      "Epoch 8/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.3413 - accuracy: 0.4272 - val_loss: 1.2774 - val_accuracy: 0.4806\n",
      "Epoch 9/50\n",
      "3443/3443 [==============================] - 2s 438us/sample - loss: 1.3137 - accuracy: 0.4246 - val_loss: 1.2563 - val_accuracy: 0.4871\n",
      "Epoch 10/50\n",
      "3443/3443 [==============================] - 2s 436us/sample - loss: 1.3006 - accuracy: 0.4453 - val_loss: 1.2203 - val_accuracy: 0.4841\n",
      "Epoch 11/50\n",
      "3443/3443 [==============================] - 2s 438us/sample - loss: 1.2827 - accuracy: 0.4522 - val_loss: 1.2175 - val_accuracy: 0.4950\n",
      "Epoch 12/50\n",
      "3443/3443 [==============================] - 2s 436us/sample - loss: 1.2844 - accuracy: 0.4389 - val_loss: 1.2066 - val_accuracy: 0.4980\n",
      "Epoch 13/50\n",
      "3443/3443 [==============================] - 2s 443us/sample - loss: 1.2479 - accuracy: 0.4548 - val_loss: 1.2008 - val_accuracy: 0.5098\n",
      "Epoch 14/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.2389 - accuracy: 0.4589 - val_loss: 1.1773 - val_accuracy: 0.5229\n",
      "Epoch 15/50\n",
      "3443/3443 [==============================] - 2s 436us/sample - loss: 1.2430 - accuracy: 0.4554 - val_loss: 1.1509 - val_accuracy: 0.5255\n",
      "Epoch 16/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.2399 - accuracy: 0.4598 - val_loss: 1.1420 - val_accuracy: 0.5260\n",
      "Epoch 17/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.2166 - accuracy: 0.4755 - val_loss: 1.1191 - val_accuracy: 0.5277\n",
      "Epoch 18/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.2226 - accuracy: 0.4592 - val_loss: 1.1297 - val_accuracy: 0.5369\n",
      "Epoch 19/50\n",
      "3443/3443 [==============================] - 2s 450us/sample - loss: 1.2100 - accuracy: 0.4740 - val_loss: 1.1124 - val_accuracy: 0.5456\n",
      "Epoch 20/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.1962 - accuracy: 0.4755 - val_loss: 1.1010 - val_accuracy: 0.5386\n",
      "Epoch 21/50\n",
      "3443/3443 [==============================] - 2s 438us/sample - loss: 1.2040 - accuracy: 0.4766 - val_loss: 1.0911 - val_accuracy: 0.5565\n",
      "Epoch 22/50\n",
      "3443/3443 [==============================] - 2s 439us/sample - loss: 1.1950 - accuracy: 0.4726 - val_loss: 1.0971 - val_accuracy: 0.5552\n",
      "Epoch 23/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.1822 - accuracy: 0.4839 - val_loss: 1.0784 - val_accuracy: 0.5648\n",
      "Epoch 24/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.1683 - accuracy: 0.4839 - val_loss: 1.0611 - val_accuracy: 0.5434\n",
      "Epoch 25/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.1660 - accuracy: 0.4993 - val_loss: 1.0653 - val_accuracy: 0.5535\n",
      "Epoch 26/50\n",
      "3443/3443 [==============================] - 2s 438us/sample - loss: 1.1623 - accuracy: 0.5048 - val_loss: 1.0449 - val_accuracy: 0.5596\n",
      "Epoch 27/50\n",
      "3443/3443 [==============================] - 2s 438us/sample - loss: 1.1411 - accuracy: 0.5135 - val_loss: 1.0315 - val_accuracy: 0.5635\n",
      "Epoch 28/50\n",
      "3443/3443 [==============================] - 2s 438us/sample - loss: 1.1314 - accuracy: 0.5144 - val_loss: 1.0254 - val_accuracy: 0.5644\n",
      "Epoch 29/50\n",
      "3443/3443 [==============================] - 2s 438us/sample - loss: 1.1354 - accuracy: 0.5103 - val_loss: 1.0255 - val_accuracy: 0.5648\n",
      "Epoch 30/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.1165 - accuracy: 0.5170 - val_loss: 1.0317 - val_accuracy: 0.5639\n",
      "Epoch 31/50\n",
      "3443/3443 [==============================] - 2s 436us/sample - loss: 1.1260 - accuracy: 0.5147 - val_loss: 1.0255 - val_accuracy: 0.5683\n",
      "Epoch 32/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.1211 - accuracy: 0.5274 - val_loss: 1.0194 - val_accuracy: 0.5840\n",
      "Epoch 33/50\n",
      "3443/3443 [==============================] - 2s 442us/sample - loss: 1.1199 - accuracy: 0.5182 - val_loss: 1.0084 - val_accuracy: 0.5827\n",
      "Epoch 34/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.1063 - accuracy: 0.5228 - val_loss: 1.0102 - val_accuracy: 0.5757\n",
      "Epoch 35/50\n",
      "3443/3443 [==============================] - 2s 438us/sample - loss: 1.1008 - accuracy: 0.5199 - val_loss: 1.0015 - val_accuracy: 0.5757\n",
      "Epoch 36/50\n",
      "3443/3443 [==============================] - 2s 436us/sample - loss: 1.1231 - accuracy: 0.5225 - val_loss: 1.0071 - val_accuracy: 0.5757\n",
      "Epoch 37/50\n",
      "3443/3443 [==============================] - 2s 436us/sample - loss: 1.1119 - accuracy: 0.5179 - val_loss: 0.9989 - val_accuracy: 0.5674\n",
      "Epoch 38/50\n",
      "3443/3443 [==============================] - 1s 435us/sample - loss: 1.1095 - accuracy: 0.5176 - val_loss: 1.0009 - val_accuracy: 0.5696\n",
      "Epoch 39/50\n",
      "3443/3443 [==============================] - 2s 446us/sample - loss: 1.1080 - accuracy: 0.5176 - val_loss: 1.0102 - val_accuracy: 0.5605\n",
      "Epoch 40/50\n",
      "3443/3443 [==============================] - 2s 436us/sample - loss: 1.1093 - accuracy: 0.5164 - val_loss: 1.0169 - val_accuracy: 0.5687\n",
      "Epoch 41/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.1010 - accuracy: 0.5243 - val_loss: 0.9980 - val_accuracy: 0.5661\n",
      "Epoch 42/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.0927 - accuracy: 0.5202 - val_loss: 0.9948 - val_accuracy: 0.5661\n",
      "Epoch 43/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.0767 - accuracy: 0.5266 - val_loss: 0.9990 - val_accuracy: 0.5858\n",
      "Epoch 44/50\n",
      "3443/3443 [==============================] - 2s 436us/sample - loss: 1.0987 - accuracy: 0.5231 - val_loss: 0.9952 - val_accuracy: 0.5696\n",
      "Epoch 45/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.0964 - accuracy: 0.5237 - val_loss: 0.9846 - val_accuracy: 0.5735\n",
      "Epoch 46/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.0892 - accuracy: 0.5245 - val_loss: 0.9939 - val_accuracy: 0.5543\n",
      "Epoch 47/50\n",
      "3443/3443 [==============================] - 2s 439us/sample - loss: 1.0888 - accuracy: 0.5228 - val_loss: 0.9904 - val_accuracy: 0.5635\n",
      "Epoch 48/50\n",
      "3443/3443 [==============================] - 2s 437us/sample - loss: 1.0744 - accuracy: 0.5274 - val_loss: 1.0043 - val_accuracy: 0.5583\n",
      "Epoch 49/50\n",
      "3443/3443 [==============================] - 2s 439us/sample - loss: 1.0779 - accuracy: 0.5222 - val_loss: 0.9926 - val_accuracy: 0.5635\n",
      "Epoch 50/50\n",
      "3443/3443 [==============================] - 2s 439us/sample - loss: 1.0718 - accuracy: 0.5237 - val_loss: 0.9809 - val_accuracy: 0.5753\n",
      "Training Accuracy: 0.6358\n",
      "Testing Accuracy:  0.5753\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(X_train, newsgroup_train.target,\n",
    "                    epochs=50,\n",
    "                    verbose=True,\n",
    "                    validation_data=(X_test, newsgroup_test.target),\n",
    "                    batch_size=10)\n",
    "loss, accuracy = model.evaluate(X_train, newsgroup_train.target, verbose=False)\n",
    "print(\"Training Accuracy: {:.4f}\".format(accuracy))\n",
    "loss, accuracy = model.evaluate(X_test, newsgroup_test.target, verbose=False)\n",
    "print(\"Testing Accuracy:  {:.4f}\".format(accuracy))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot training & validation accuracy values\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "# Plot training & validation loss values\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('Model loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}