2BwD-Layer-DK-AlignMNiST-5.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-04-20T21:52:04.548207",
     "start_time": "2016-04-20T21:52:04.544771"
    }
   },
   "source": [
    "## 2 Layer MLP for Dark Knoweldge Experiments w/AlignMNIST\n",
    "\n",
    "75 epochs of RMSProp on AlignMNIST\n",
    "\n",
    "800 x 800 network with normalized the weights after each epoch\n",
    "\n",
    "no dropout .. another baseline\n",
    "\n",
    "\n",
    "see  http://www.r2rt.com/posts/implementations/2016-03-29-implementing-batch-normalization-tensorflow/\n",
    "\n",
    "https://github.com/tensorflow/tensorflow/blob/b3621c95160a916d4d255f9f44318b9d465701af/tensorflow/contrib/layers/python/layers/layers.py\n",
    "\n",
    "\n",
    "https://www.reddit.com/r/MachineLearning/comments/2bopxs/question_about_the_maxnorm_constraint_used_with/\n",
    "\n",
    "\n",
    "\n",
    "https://arxiv.org/pdf/1207.0580.pdf\n",
    "\n",
    "https://arxiv.org/pdf/1503.02531v1.pdf\n",
    "\n",
    "http://www.kdnuggets.com/2015/04/preventing-overfitting-neural-networks.html\n",
    "\n",
    "https://www.reddit.com/r/MachineLearning/comments/2bopxs/question_about_the_maxnorm_constraint_used_with/\n",
    "\n",
    "and https://github.com/tensorflow/tensorflow/issues/608"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T22:37:57.602940",
     "start_time": "2016-06-16T22:37:55.723796"
    },
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T22:37:57.628912",
     "start_time": "2016-06-16T22:37:57.608540"
    },
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "%run augmentmnist.py"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T22:37:57.635100",
     "start_time": "2016-06-16T22:37:57.631672"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "num_epochs = 75\n",
    "batch_size = 125\n",
    "\n",
    "learning_rate = 0.001\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T22:37:57.643958",
     "start_time": "2016-06-16T22:37:57.637188"
    },
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Network Parameters\n",
    "n_hidden_1 = 800 # 1st layer num features\n",
    "n_hidden_2 = 800 # 2nd layer num features\n",
    "n_input = 784 # MNIST data input (img shape: 28*28)\n",
    "n_classes = 10 # MNIST total classes (0-9 digits)\n",
    "\n",
    "std_0  = np.sqrt(6.0/(n_input+n_hidden_1))\n",
    "std_h1 = np.sqrt(6.0/(n_hidden_1+n_hidden_2))\n",
    "std_h2 = np.sqrt(6.0/(n_hidden_2+n_classes))\n",
    "\n",
    "logfile = \"2BwD-layer-dlk-alignmnist-4\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T22:37:57.652907",
     "start_time": "2016-06-16T22:37:57.645573"
    },
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# tf Graph input\n",
    "x = tf.placeholder(\"float\", [None, n_input])\n",
    "y = tf.placeholder(\"float\", [None, n_classes])\n",
    "\n",
    "xt = tf.placeholder(\"float\", [None, n_input])\n",
    "yt = tf.placeholder(\"float\", [None, n_classes])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### need to try again with Dropout on both layers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T22:37:58.526181",
     "start_time": "2016-06-16T22:37:58.522441"
    },
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def multilayer_perceptron(_X, _weights, _biases):\n",
    "    layer_1 = tf.nn.relu(tf.add(tf.matmul(_X, _weights['h1']), _biases['b1']))\n",
    "    layer_2 =tf.nn.relu(tf.add(tf.matmul(layer_1, _weights['h2']), _biases['b2']))\n",
    "    return tf.matmul(layer_2, weights['out']) + biases['out']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T22:37:58.680867",
     "start_time": "2016-06-16T22:37:58.676985"
    },
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def multilayer_perceptron_test(_X, _weights, _biases):\n",
    "    layer_1 = tf.nn.relu(tf.add(tf.matmul(_X, _weights['h1']), _biases['b1']))\n",
    "    layer_2 = tf.nn.relu(tf.add(tf.matmul(layer_1, _weights['h2']), _biases['b2']))\n",
    "    return tf.matmul(layer_2, weights['out']) + biases['out']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T22:37:58.880030",
     "start_time": "2016-06-16T22:37:58.843497"
    },
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Store layers weight & bias\n",
    "weights = {\n",
    "    'h1': tf.Variable(tf.random_normal([n_input, n_hidden_1], stddev=std_0)),\n",
    "    'h2': tf.Variable(tf.random_normal([n_hidden_1, n_hidden_2], stddev=std_h1)),\n",
    "    'out': tf.Variable(tf.random_normal([n_hidden_2, n_classes], stddev=std_h2))\n",
    "}\n",
    "biases = {\n",
    "    'b1': tf.Variable(tf.random_normal([n_hidden_1], stddev=0.1)),\n",
    "    'b2': tf.Variable(tf.random_normal([n_hidden_2], stddev=0.01)),\n",
    "    'out': tf.Variable(tf.random_normal([n_classes], stddev=0.001))\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T22:37:59.536509",
     "start_time": "2016-06-16T22:37:59.525239"
    },
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "mlp = multilayer_perceptron(x, weights, biases )\n",
    "mlp_test = multilayer_perceptron_test(xt, weights, biases  )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T22:38:00.029439",
     "start_time": "2016-06-16T22:38:00.018054"
    },
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "cost =  tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(mlp, y)) \n",
    "gs = tf.get_variable(\"global_step\",[],trainable=False,initializer=tf.constant_initializer(0))\n",
    "lr = tf.constant(learning_rate) "
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T15:33:44.245778",
     "start_time": "2016-06-16T15:33:44.109400"
    },
    "collapsed": false,
    "scrolled": false
   },
   "source": [
    "train_op = tf.contrib.layers.optimize_loss(cost, global_step=gs, learning_rate=lr,optimizer=\"RMSProp\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-15T23:46:29.570951",
     "start_time": "2016-06-15T23:46:29.568744"
    }
   },
   "source": [
    "### don't include dropout in accuracy calculations"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T15:33:45.106743",
     "start_time": "2016-06-16T15:33:45.085919"
    },
    "collapsed": false
   },
   "source": [
    "#with tf.name_scope(\"training accuracy\"):\n",
    "pred = tf.equal(tf.argmax(mlp, 1), tf.argmax(y, 1)) # Count correct predictions\n",
    "train_acc_op = tf.reduce_mean(tf.cast(pred, \"float\"))  # Cast boolean to float to average\n",
    "tf.scalar_summary(\"training accuracy\", train_acc_op)"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T15:33:46.093438",
     "start_time": "2016-06-16T15:33:46.081762"
    },
    "collapsed": false
   },
   "source": [
    "test_pred = tf.equal(tf.argmax(mlp_test, 1), tf.argmax(yt, 1)) # Count correct predictions\n",
    "test_acc_op = tf.reduce_mean(tf.cast(test_pred, \"float\"))  # Cast boolean to float to average\n",
    "tf.scalar_summary(\"test 0 accuracy\", test_acc_op)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T22:38:15.361463",
     "start_time": "2016-06-16T22:38:15.129739"
    },
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1m\u001b[36m2-layer-dk-alignmnist\u001b[m\u001b[m           \u001b[1m\u001b[36m2-layer-modern-mlp-alignmnist\u001b[m\u001b[m\r\n",
      "\u001b[1m\u001b[36m2-layer-dlk-alignmnist\u001b[m\u001b[m          \u001b[1m\u001b[36m2BwD-layer-dlk-alignmnist\u001b[m\u001b[m\r\n",
      "\u001b[1m\u001b[36m2-layer-mlp-alignmnist-wdropout\u001b[m\u001b[m \u001b[1m\u001b[36m5-layer-mlp-alignmnist-wdropout\u001b[m\u001b[m\r\n",
      "\u001b[1m\u001b[36m2-layer-mlp-mnist-temp\u001b[m\u001b[m          \u001b[1m\u001b[36m5-layer-mlp-infimnist\u001b[m\u001b[m\r\n",
      "\u001b[1m\u001b[36m2-layer-mlp-mnist-watch\u001b[m\u001b[m\r\n"
     ]
    }
   ],
   "source": [
    "!rm -rf ./logs/{logfile}\n",
    "!ls logs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-05-04T21:57:37.534602",
     "start_time": "2016-05-04T21:57:37.532739"
    },
    "collapsed": true
   },
   "source": [
    "### Original MNIST Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T22:38:16.932266",
     "start_time": "2016-06-16T22:38:16.092793"
    },
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting ./train-images-idx3-ubyte.gz\n",
      "Extracting ./train-labels-idx1-ubyte.gz\n",
      "Extracting ./t10k-images-idx3-ubyte.gz\n",
      "Extracting ./t10k-labels-idx1-ubyte.gz\n"
     ]
    }
   ],
   "source": [
    "from tensorflow.examples.tutorials.mnist import input_data\n",
    "mnist = input_data.read_data_sets(\".\", one_hot=True)\n",
    "trX_0, trY_0 = mnist.train.images, mnist.train.labels\n",
    "teX_0, teY_0 = mnist.test.images, mnist.test.labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T22:38:17.617307",
     "start_time": "2016-06-16T22:38:17.612219"
    },
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "alignmnist = AlignMNIST()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-14T18:12:43.510102",
     "start_time": "2016-06-14T18:12:43.502711"
    }
   },
   "source": [
    " see https://github.com/tensorflow/tensorflow/issues/424"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T22:38:18.678973",
     "start_time": "2016-06-16T22:38:18.674926"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from tensorflow.python.framework import ops\n",
    "from tensorflow.python.ops import array_ops\n",
    "from tensorflow.python.ops import control_flow_ops\n",
    "from tensorflow.python.ops import math_ops\n",
    "\n",
    "from tensorflow.python.ops import variable_scope as vs\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T22:38:19.711451",
     "start_time": "2016-06-16T22:38:19.673352"
    },
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "# do i have this right  ?\n",
    "\n",
    "def max_norm_constraint(_weights, _biases):\n",
    "    \n",
    "    norms0 = np.array([50.0, 60.0, 6.0, 3.0, 1.0, 1.0])\n",
    "    norms0 = norms0*1.0\n",
    "\n",
    "    n0 = tf.constant(norms0[0], dtype=tf.float32)\n",
    "    n1 = tf.constant(norms0[1],  dtype=tf.float32)\n",
    "    n2 = tf.constant(norms0[2],  dtype=tf.float32)\n",
    "    n3 = tf.constant(norms0[3],  dtype=tf.float32)\n",
    "    n4 = tf.constant(norms0[4],  dtype=tf.float32)\n",
    "    n5 = tf.constant(norms0[5],  dtype=tf.float32)\n",
    "    \n",
    "    norm_h1 =   tf.sqrt(tf.reduce_sum(tf.square(_weights['h1'])))\n",
    "    norm_h2 =   tf.sqrt(tf.reduce_sum(tf.square(_weights['h2'])))\n",
    "    norm_hout = tf.sqrt(tf.reduce_sum(tf.square(_weights['out'])))\n",
    "                        \n",
    "    norm_b1 =   tf.sqrt(tf.reduce_sum(tf.square(_biases['b1'])))\n",
    "    norm_b2 =   tf.sqrt(tf.reduce_sum(tf.square(_biases['b2'])))\n",
    "    norm_bout = tf.sqrt(tf.reduce_sum(tf.square(_biases['out'])))\n",
    "\n",
    "    \n",
    "    pred = tf.greater(norm_h1 , n0)\n",
    "    _weights['h1'] = control_flow_ops.cond(pred, lambda: tf.div(_weights['h1'], norm_h1), lambda: _weights['h1'])\n",
    "    \n",
    "    pred = tf.greater(norm_h2 , n1)\n",
    "    _weights['h2'] = control_flow_ops.cond(pred, lambda: tf.div(_weights['h2'], norm_h2), lambda: _weights['h2'])\n",
    "    \n",
    "    pred = tf.greater(norm_hout , n2)\n",
    "    _weights['out'] = control_flow_ops.cond(pred, lambda: tf.div(_weights['out'], norm_hout), lambda: _weights['out'])\n",
    "            \n",
    "    pred = tf.greater(norm_b1 , n3)\n",
    "    _biases['b1'] = control_flow_ops.cond(pred, lambda: tf.div(_biases['b1'], norm_b1), lambda: _biases['b1'])\n",
    "    \n",
    "    pred = tf.greater(norm_b2 , n4)\n",
    "    _biases['b2'] = control_flow_ops.cond(pred, lambda: tf.div(_biases['b2'], norm_b2), lambda: _biases['b2'])\n",
    "\n",
    "    pred = tf.greater(norm_bout , n5)\n",
    "    _biases['out'] = control_flow_ops.cond(pred, lambda: tf.div(_biases['out'], norm_bout), lambda: _biases['out'])\n",
    "       \n",
    "    return tf.constant(norms0[0], dtype=tf.float32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T15:33:51.063945",
     "start_time": "2016-06-16T15:33:51.003699"
    },
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-17T08:40:16.507359",
     "start_time": "2016-06-17T06:32:35.619442"
    },
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(0, 0.96998334, 0.95929998)\n",
      "[1.000017, 50.92754, 0.9999997, 2.8767266, 0.38060847, 0.019276755]\n",
      "(1, 0.95885003, 0.97460002)\n",
      "[1.0000167, 53.540356, 1.0000001, 2.8934255, 0.61607522, 0.066759184]\n",
      "(0, 0.97486669, 0.96749997)\n",
      "[49.990677, 50.907757, 1.0, 2.756165, 0.37869224, 0.022361267]\n",
      "(1, 0.96385002, 0.9777)\n",
      "[1.0000156, 53.527569, 1.0000001, 2.7622385, 0.62265515, 0.057664622]\n",
      "(0, 0.97235, 0.96579999)\n",
      "[50.000713, 50.935467, 0.99999982, 2.7350247, 0.36257568, 0.017452713]\n",
      "(1, 0.9684, 0.97729999)\n",
      "[1.0000184, 53.530972, 0.99999982, 2.7530963, 0.55639946, 0.051071223]\n",
      "(0, 0.96885002, 0.95920002)\n",
      "[49.995628, 50.953915, 0.99999994, 2.9125488, 0.37396094, 0.024777452]\n",
      "(1, 0.96641666, 0.97780001)\n",
      "[1.0000163, 53.568703, 1.0, 2.9249277, 0.61312139, 0.058107249]\n",
      "(0, 0.97451669, 0.96670002)\n",
      "[1.0000157, 50.924599, 0.99999976, 2.8730028, 0.3683621, 0.027475618]\n",
      "(1, 0.96945, 0.98269999)\n",
      "[1.0000178, 53.530125, 1.0000001, 2.8856587, 0.60122418, 0.064867288]\n",
      "(0, 0.97065002, 0.96420002)\n",
      "[1.0000157, 50.822449, 0.9999997, 2.7012658, 0.35330543, 0.020614164]\n",
      "(1, 0.95921665, 0.9745)\n",
      "[1.0000151, 53.420876, 1.0000001, 2.7137015, 0.56904775, 0.053243078]\n",
      "(0, 0.96773332, 0.95990002)\n",
      "[1.0000153, 50.826626, 1.0, 2.7940764, 0.36335537, 0.027420137]\n",
      "(1, 0.96364999, 0.97399998)\n",
      "[1.0000179, 53.42614, 1.0000002, 2.803489, 0.58510596, 0.054764029]\n",
      "(0, 0.9752, 0.97000003)\n",
      "[1.0000162, 50.890965, 0.99999976, 2.8695803, 0.3540231, 0.02699006]\n",
      "(1, 0.96885002, 0.97970003)\n",
      "[1.0000157, 53.51622, 1.0000002, 2.8780589, 0.58065963, 0.062361676]\n",
      "(0, 0.96591669, 0.96039999)\n",
      "[1.0000193, 50.918381, 1.0000001, 2.778276, 0.35312372, 0.022264818]\n",
      "(1, 0.96385002, 0.97640002)\n",
      "[1.0000145, 53.48494, 0.99999946, 2.7888839, 0.56666797, 0.055342883]\n",
      "(0, 0.97443336, 0.96740001)\n",
      "[49.980507, 50.862331, 1.0, 2.7979827, 0.35732207, 0.023671536]\n",
      "(1, 0.96476668, 0.97430003)\n",
      "[1.0000168, 53.454739, 1.0, 2.8002858, 0.58212137, 0.049010027]\n",
      "(0, 0.97191668, 0.96380001)\n",
      "[49.929234, 50.880993, 0.99999988, 2.8428619, 0.35699555, 0.026017824]\n",
      "(1, 0.96653336, 0.97619998)\n",
      "[1.0000167, 53.470264, 0.99999976, 2.8537741, 0.58003551, 0.055631515]\n",
      "(0, 0.97406667, 0.96759999)\n",
      "[1.0000166, 50.940628, 1.0000001, 2.837714, 0.36658946, 0.029240193]\n",
      "(1, 0.96539998, 0.97970003)\n",
      "[1.0000169, 53.524902, 1.0, 2.8426883, 0.61969852, 0.065458141]\n",
      "(0, 0.9544, 0.94870001)\n",
      "[1.0000148, 50.87191, 1.0, 2.8817334, 0.3841919, 0.033071924]\n",
      "(1, 0.9497, 0.96609998)\n",
      "[1.0000148, 53.498466, 1.0, 2.8977497, 0.59992927, 0.057555925]\n",
      "(0, 0.96563333, 0.95700002)\n",
      "[1.0000153, 50.861198, 1.0000001, 2.8487859, 0.36239704, 0.032241065]\n",
      "(1, 0.96210003, 0.97310001)\n",
      "[1.000016, 53.50423, 0.99999988, 2.8605552, 0.59835184, 0.065564066]\n",
      "(0, 0.97418332, 0.96780002)\n",
      "[1.0000198, 50.93124, 1.0000001, 2.8304467, 0.35196421, 0.02605392]\n",
      "(1, 0.96025002, 0.97460002)\n",
      "[1.0000172, 53.567638, 0.99999982, 2.8371611, 0.57707393, 0.059184026]\n",
      "(0, 0.97235, 0.9648)\n",
      "[1.0000142, 50.952728, 1.0000001, 2.8853395, 0.34566087, 0.025663774]\n",
      "(1, 0.95411664, 0.96850002)\n",
      "[1.0000173, 53.591976, 1.0000004, 2.8935962, 0.5675894, 0.048138615]\n",
      "(0, 0.96918333, 0.9623)\n",
      "[1.0000162, 50.881264, 1.0000001, 2.7930098, 0.37094587, 0.016652042]\n",
      "(1, 0.96103334, 0.97149998)\n",
      "[1.0000137, 53.490391, 1.0, 2.8063152, 0.60702485, 0.057617817]\n",
      "(0, 0.9745, 0.96789998)\n",
      "[1.0000149, 50.91605, 1.0, 2.7885134, 0.36343342, 0.023819007]\n",
      "(1, 0.96531665, 0.97829998)\n",
      "[1.0000141, 53.522068, 1.0000002, 2.789418, 0.60630041, 0.051591542]\n",
      "(0, 0.96810001, 0.96249998)\n",
      "[49.977612, 50.934597, 1.0000001, 2.8805339, 0.35472098, 0.018664306]\n",
      "(1, 0.96566665, 0.97600001)\n",
      "[1.0000216, 53.560612, 0.99999982, 2.8906672, 0.55360711, 0.053096704]\n",
      "(0, 0.97098333, 0.96579999)\n",
      "[1.0000155, 50.932449, 1.0, 2.6607754, 0.36634141, 0.022368256]\n",
      "(1, 0.96673334, 0.98079997)\n",
      "[1.0000167, 53.538357, 1.0000002, 2.6815419, 0.58903372, 0.056001782]\n",
      "(0, 0.97491664, 0.9673)\n",
      "[49.985912, 50.901463, 1.0000001, 2.8831728, 0.37138125, 0.022448597]\n",
      "(1, 0.96695, 0.97689998)\n",
      "[1.0000165, 53.500172, 1.0000002, 2.8981783, 0.6084767, 0.059165452]\n",
      "(0, 0.97496665, 0.96679997)\n",
      "[1.0000175, 50.911713, 1.0000001, 2.696403, 0.36073729, 0.021696063]\n",
      "(1, 0.95829999, 0.9763)\n",
      "[1.000016, 53.511841, 0.99999982, 2.7049317, 0.57463193, 0.053143103]\n",
      "(0, 0.96903336, 0.96429998)\n",
      "[50.000717, 50.962067, 0.99999994, 2.8298428, 0.37987405, 0.027217451]\n",
      "(1, 0.96136665, 0.97350001)\n",
      "[1.0000172, 53.565346, 1.0000002, 2.826628, 0.58220357, 0.0459578]\n",
      "(0, 0.97521669, 0.96829998)\n",
      "[1.000015, 50.889957, 0.9999997, 2.7468908, 0.36910588, 0.022478014]\n",
      "(1, 0.9641, 0.97850001)\n",
      "[1.000017, 53.442341, 0.99999988, 2.7517607, 0.57570773, 0.046970081]\n",
      "(0, 0.97183335, 0.96609998)\n",
      "[1.000016, 50.879093, 1.0000001, 2.8434317, 0.35995436, 0.027257562]\n",
      "(1, 0.95708334, 0.97399998)\n",
      "[1.0000128, 53.525089, 1.0000001, 2.8466785, 0.57279336, 0.061108958]\n",
      "(0, 0.96773332, 0.96079999)\n",
      "[1.0000176, 50.883778, 0.99999964, 2.8179605, 0.36485848, 0.021355283]\n",
      "(1, 0.96481669, 0.97890002)\n",
      "[1.0000176, 53.494869, 1.0000001, 2.8311577, 0.60233796, 0.054606881]\n",
      "(0, 0.97081667, 0.96469998)\n",
      "[49.906662, 50.892262, 0.99999994, 2.8900545, 0.36061263, 0.029162658]\n",
      "(1, 0.96476668, 0.97610003)\n",
      "[1.000015, 53.523991, 1.0000001, 2.9018588, 0.60438496, 0.063065626]\n",
      "(0, 0.97558331, 0.96700001)\n",
      "[49.927505, 50.899693, 1.0, 2.8644547, 0.359759, 0.029575339]\n",
      "(1, 0.96451664, 0.97589999)\n",
      "[1.000017, 53.473709, 0.9999997, 2.8914728, 0.59150791, 0.059912868]\n",
      "(0, 0.97140002, 0.9641)\n",
      "[49.984074, 50.841862, 0.99999976, 2.7242973, 0.35774747, 0.0076253214]\n",
      "(1, 0.96925002, 0.97970003)\n",
      "[1.0000154, 53.424614, 1.0, 2.7476628, 0.57087117, 0.050241761]\n",
      "(0, 0.97095001, 0.9666)\n",
      "[1.0000132, 50.915916, 0.99999994, 2.7971356, 0.36399084, 0.022653092]\n",
      "(1, 0.96748334, 0.97729999)\n",
      "[1.0000161, 53.516445, 0.99999988, 2.8162808, 0.58980477, 0.055751204]\n",
      "(0, 0.95674998, 0.94859999)\n",
      "[49.985992, 50.92585, 0.99999982, 2.8105903, 0.37502277, 0.026711127]\n",
      "(1, 0.9612667, 0.97549999)\n",
      "[1.0000149, 53.536121, 0.99999994, 2.8191445, 0.577793, 0.050797157]\n",
      "(0, 0.96735001, 0.95920002)\n",
      "[49.992851, 50.862564, 1.0000001, 2.7397482, 0.34986478, 0.026597612]\n",
      "(1, 0.95738333, 0.97039998)\n",
      "[1.0000147, 53.490227, 1.0, 2.7560818, 0.56476682, 0.062980898]\n",
      "(0, 0.95291668, 0.94340003)\n",
      "[1.0000155, 50.877838, 1.0000001, 2.7630894, 0.37253338, 0.022718973]\n",
      "(1, 0.95091665, 0.96850002)\n",
      "[1.0000165, 53.489517, 0.99999982, 2.7894466, 0.58475953, 0.048442643]\n",
      "(0, 0.97718334, 0.97229999)\n",
      "[49.974445, 50.869312, 1.0000002, 2.7525833, 0.37191471, 0.021477688]\n",
      "(1, 0.96128333, 0.9756)\n",
      "[1.0000142, 53.461384, 1.0000001, 2.7744544, 0.61099648, 0.052689403]\n",
      "(0, 0.97598332, 0.96799999)\n",
      "[1.0000167, 50.920368, 0.99999976, 2.9184501, 0.37166908, 0.023583716]\n",
      "(1, 0.96153331, 0.97430003)\n",
      "[1.0000181, 53.529381, 0.99999976, 2.9235458, 0.61074197, 0.058614213]\n",
      "(0, 0.97381669, 0.96719998)\n",
      "[1.0000149, 50.977772, 0.99999982, 2.8272948, 0.36355981, 0.026267249]\n",
      "(1, 0.95486665, 0.97280002)\n",
      "[1.0000143, 53.526264, 0.99999994, 2.8180759, 0.60059142, 0.059196189]\n",
      "(0, 0.96876669, 0.96090001)\n",
      "[1.000017, 50.938091, 1.0, 2.8796911, 0.37050515, 0.031917367]\n",
      "(1, 0.96569997, 0.97729999)\n",
      "[1.0000162, 53.503494, 1.0, 2.8864105, 0.58848524, 0.065806188]\n",
      "(0, 0.96708333, 0.96270001)\n",
      "[49.955547, 50.835979, 0.99999988, 2.9078405, 0.3611663, 0.017487355]\n",
      "(1, 0.96681666, 0.97719997)\n",
      "[1.0000175, 53.409416, 0.99999994, 2.9146433, 0.61665529, 0.051537361]\n",
      "(0, 0.97318333, 0.96810001)\n",
      "[1.0000175, 50.870468, 0.99999994, 2.7386017, 0.36191359, 0.023741189]\n",
      "(1, 0.96853334, 0.97909999)\n",
      "[1.0000168, 53.467007, 1.0000001, 2.7594972, 0.57029706, 0.058577072]\n",
      "(0, 0.97783333, 0.97180003)\n",
      "[1.0000159, 50.917858, 1.0000001, 2.9282799, 0.36095461, 0.020651452]\n",
      "(1, 0.96601665, 0.97649997)\n",
      "[1.0000186, 53.536434, 1.0000001, 2.9269891, 0.59945667, 0.056724001]\n",
      "(0, 0.97251666, 0.96329999)\n",
      "[1.0000162, 50.877087, 1.0000001, 2.7225003, 0.38237503, 0.031053251]\n",
      "(1, 0.95730001, 0.97369999)\n",
      "[1.0000182, 53.47863, 1.0000002, 2.7281346, 0.60203934, 0.070472933]\n",
      "(0, 0.97323334, 0.96640003)\n",
      "[1.0000173, 50.934605, 0.99999994, 2.7881818, 0.37040237, 0.028279265]\n",
      "(1, 0.96890002, 0.97799999)\n",
      "[1.0000172, 53.589817, 1.0000004, 2.7915542, 0.58258343, 0.063178517]\n",
      "(0, 0.97404999, 0.9677)\n",
      "[49.953739, 50.960316, 1.0000001, 2.787766, 0.38867322, 0.030657448]\n",
      "(1, 0.96318334, 0.97670001)\n",
      "[1.0000131, 53.57893, 0.99999994, 2.7971542, 0.62398666, 0.061247937]\n",
      "(0, 0.97513336, 0.96850002)\n",
      "[1.0000161, 50.89011, 1.0, 2.8418999, 0.37231997, 0.021632401]\n",
      "(1, 0.96039999, 0.97579998)\n",
      "[1.0000156, 53.489719, 1.0, 2.866663, 0.60695904, 0.066096053]\n",
      "(0, 0.96271664, 0.95670003)\n",
      "[49.923473, 50.904026, 1.0000001, 2.8243499, 0.36785084, 0.021125719]\n",
      "(1, 0.96160001, 0.97250003)\n",
      "[1.0000161, 53.533859, 1.0000001, 2.8232355, 0.57265991, 0.057187218]\n",
      "(0, 0.96413332, 0.95289999)\n",
      "[1.0000185, 50.821026, 1.0000002, 2.7788675, 0.36849761, 0.029428545]\n",
      "(1, 0.96921664, 0.9777)\n",
      "[1.0000176, 53.412075, 1.0, 2.8003159, 0.58152008, 0.058929592]\n",
      "(0, 0.97231668, 0.96649998)\n",
      "[1.0000151, 50.889832, 0.99999988, 2.9510243, 0.36465535, 0.017621417]\n",
      "(1, 0.95735002, 0.97049999)\n",
      "[1.0000135, 53.498425, 0.99999994, 2.9649789, 0.59981912, 0.067257598]\n",
      "(0, 0.9740833, 0.96719998)\n",
      "[1.0000167, 50.785908, 1.0000002, 2.9305594, 0.36295956, 0.021150036]\n",
      "(1, 0.95676666, 0.97219998)\n",
      "[1.0000155, 53.367298, 1.0000001, 2.9451911, 0.5893718, 0.056507189]\n",
      "(0, 0.97610003, 0.96820003)\n",
      "[49.97353, 50.863682, 1.0000001, 2.7723937, 0.36108696, 0.023100434]\n",
      "(1, 0.96328336, 0.97509998)\n",
      "[1.0000155, 53.448528, 1.0000002, 2.7824221, 0.59408885, 0.056548048]\n",
      "(0, 0.96813333, 0.96030003)\n",
      "[49.967266, 50.874104, 0.99999976, 2.9095025, 0.3614794, 0.035430372]\n",
      "(1, 0.9683333, 0.98110002)\n",
      "[1.000015, 53.49934, 0.99999988, 2.9186344, 0.59742528, 0.072346315]\n",
      "(0, 0.97219998, 0.96560001)\n",
      "[1.0000162, 50.878876, 1.0, 2.816819, 0.35387453, 0.025757741]\n",
      "(1, 0.96354997, 0.97589999)\n",
      "[1.0000178, 53.490704, 0.99999964, 2.835778, 0.56728077, 0.060836762]\n",
      "(0, 0.96638334, 0.96020001)\n",
      "[1.0000169, 50.891029, 0.99999976, 2.6713991, 0.35211638, 0.026298309]\n",
      "(1, 0.96224999, 0.97549999)\n",
      "[1.0000176, 53.501492, 1.0, 2.6756582, 0.60063004, 0.060748391]\n",
      "(0, 0.97318333, 0.96759999)\n",
      "[49.977493, 50.920033, 0.99999994, 2.7713215, 0.3617073, 0.024535788]\n",
      "(1, 0.96539998, 0.97750002)\n",
      "[1.0000157, 53.538815, 1.0, 2.7713721, 0.58937758, 0.055804461]\n",
      "(0, 0.97681665, 0.96899998)\n",
      "[49.954311, 50.797672, 0.9999997, 2.8138187, 0.37872964, 0.02992647]\n",
      "(1, 0.96139997, 0.97649997)\n",
      "[1.0000149, 53.397293, 1.0000001, 2.8138897, 0.59850585, 0.061652254]\n",
      "(0, 0.96920002, 0.95950001)\n",
      "[1.0000139, 50.902027, 1.0000001, 2.8610458, 0.3576076, 0.014505114]\n",
      "(1, 0.96286666, 0.97530001)\n",
      "[1.000016, 53.505531, 0.99999994, 2.8787758, 0.57704365, 0.048160248]\n",
      "(0, 0.97716665, 0.96990001)\n",
      "[1.0000157, 50.920975, 0.99999976, 2.8641453, 0.35728678, 0.020333992]\n",
      "(1, 0.96845001, 0.97780001)\n",
      "[1.000018, 53.531898, 1.0, 2.8567059, 0.58086324, 0.051893715]\n",
      "(0, 0.96824998, 0.96020001)\n",
      "[1.0000148, 50.907814, 0.99999994, 2.8126602, 0.35511994, 0.021666067]\n",
      "(1, 0.96601665, 0.97820002)\n",
      "[1.0000188, 53.545818, 1.0000004, 2.8079805, 0.57984459, 0.057331394]\n",
      "(0, 0.97764999, 0.97119999)\n",
      "[49.983143, 50.92086, 1.0, 2.8133657, 0.36928329, 0.022930495]\n",
      "(1, 0.95963335, 0.97479999)\n",
      "[1.0000178, 53.562847, 0.99999994, 2.8172889, 0.60601461, 0.050179932]\n",
      "(0, 0.97413331, 0.96969998)\n",
      "[1.000016, 50.900726, 0.99999988, 2.8490202, 0.38402352, 0.024221949]\n",
      "(1, 0.95411664, 0.9727)\n",
      "[1.0000181, 53.477547, 1.0000001, 2.8704114, 0.63270819, 0.062007204]\n",
      "(0, 0.96095002, 0.94989997)\n",
      "[1.0000181, 50.949444, 0.99999982, 2.872268, 0.37548143, 0.027257603]\n",
      "(1, 0.96228331, 0.97850001)\n",
      "[1.0000156, 53.594433, 1.0000002, 2.8853838, 0.6095472, 0.06075608]\n",
      "(0, 0.97444999, 0.96630001)\n",
      "[1.000016, 50.925919, 0.99999994, 2.8095558, 0.36373505, 0.02117037]\n",
      "(1, 0.96871668, 0.98180002)\n",
      "[1.0000172, 53.553398, 1.0000001, 2.8152673, 0.60877967, 0.054705754]\n",
      "(0, 0.97253335, 0.96710002)\n",
      "[50.000401, 50.896919, 1.0000001, 2.7582185, 0.35520092, 0.024661336]\n",
      "(1, 0.96031666, 0.97329998)\n",
      "[1.000015, 53.467052, 1.0, 2.7502003, 0.56460899, 0.054221965]\n",
      "(0, 0.97686666, 0.96789998)\n",
      "[1.0000199, 50.956444, 1.0000001, 2.9214847, 0.36923987, 0.028299756]\n",
      "(1, 0.96574998, 0.97829998)\n",
      "[1.0000162, 53.556625, 0.99999976, 2.9351003, 0.61112046, 0.06513875]\n",
      "(0, 0.95773333, 0.95190001)\n",
      "[49.982063, 50.951935, 1.0, 2.8766119, 0.37477368, 0.017465597]\n",
      "(1, 0.96621668, 0.97680002)\n",
      "[1.0000161, 53.589352, 1.0000001, 2.8993292, 0.61447769, 0.057711288]\n",
      "(0, 0.97276664, 0.96560001)\n",
      "[1.0000155, 50.907948, 1.0000001, 2.7727158, 0.35719371, 0.013504178]\n",
      "(1, 0.95398331, 0.96929997)\n",
      "[1.0000141, 53.535816, 0.99999988, 2.7837546, 0.57549757, 0.053287953]\n",
      "(0, 0.9659, 0.95279998)\n",
      "[1.0000144, 50.791702, 1.0, 2.7702227, 0.36507368, 0.026313944]\n",
      "(1, 0.96573335, 0.9781)\n",
      "[1.0000142, 53.441376, 0.99999976, 2.7996135, 0.57351923, 0.056863114]\n",
      "(0, 0.96676666, 0.9562)\n",
      "[49.977379, 50.896122, 1.0000001, 2.6740358, 0.35973394, 0.033264384]\n",
      "(1, 0.96856666, 0.98019999)\n",
      "[1.0000168, 53.500027, 0.99999994, 2.6884382, 0.6053192, 0.057316579]\n",
      "(0, 0.97409999, 0.96759999)\n",
      "[49.969395, 50.965839, 0.99999982, 2.8945258, 0.38764057, 0.02953021]\n",
      "(1, 0.95756668, 0.9709)\n",
      "[1.0000172, 53.591854, 1.0000001, 2.8954461, 0.59503937, 0.060068466]\n",
      "(0, 0.96579999, 0.96039999)\n",
      "[1.0000163, 50.908932, 0.99999994, 2.8290977, 0.37228563, 0.031092858]\n",
      "(1, 0.96196669, 0.97530001)\n",
      "[1.0000154, 53.551823, 1.0, 2.8361523, 0.61943662, 0.06925603]\n",
      "(0, 0.97783333, 0.972)\n",
      "[49.958477, 50.911438, 0.99999976, 2.8338566, 0.36484751, 0.024191955]\n",
      "(1, 0.96789998, 0.98030001)\n",
      "[1.0000178, 53.4846, 0.99999982, 2.8413944, 0.6107111, 0.06164379]\n",
      "(0, 0.9738, 0.96569997)\n",
      "[49.981739, 50.889019, 0.99999994, 2.764425, 0.33854777, 0.026435502]\n",
      "(1, 0.96565002, 0.97920001)\n",
      "[1.0000162, 53.475494, 1.0000001, 2.7915831, 0.56703418, 0.056146566]\n",
      "(0, 0.96985, 0.9637)\n",
      "[1.0000166, 50.909264, 1.0000001, 2.7930882, 0.38082057, 0.03039542]\n",
      "(1, 0.96348333, 0.9738)\n",
      "[1.0000169, 53.54332, 1.0000001, 2.8047705, 0.59264946, 0.05726219]\n",
      "(0, 0.97606665, 0.96850002)\n",
      "[1.0000157, 50.879181, 1.0, 2.6625898, 0.35516223, 0.022289883]\n",
      "(1, 0.96743333, 0.97670001)\n",
      "[1.0000166, 53.51709, 0.99999988, 2.6704397, 0.56315482, 0.041939285]\n",
      "(0, 0.97358334, 0.96600002)\n",
      "[49.938293, 50.899349, 0.99999988, 2.8045113, 0.36469039, 0.031501845]\n",
      "(1, 0.96441668, 0.97539997)\n",
      "[1.0000176, 53.503609, 1.0000002, 2.8134789, 0.59520084, 0.059012093]\n",
      "(0, 0.97416669, 0.96509999)\n",
      "[49.934299, 50.902393, 0.99999946, 2.8538051, 0.36291036, 0.016406849]\n",
      "(1, 0.96155, 0.97570002)\n",
      "[1.0000138, 53.519924, 1.0000001, 2.8623972, 0.58922619, 0.054085795]\n",
      "(0, 0.97673333, 0.96799999)\n",
      "[49.98148, 50.8428, 0.99999988, 2.871573, 0.37053764, 0.031925946]\n",
      "(1, 0.95381665, 0.9734)\n",
      "[1.0000167, 53.441353, 0.99999976, 2.8924716, 0.58567637, 0.062839024]\n",
      "(0, 0.97479999, 0.96679997)\n",
      "[49.973293, 50.91021, 1.0000002, 2.8224945, 0.37661898, 0.037486248]\n",
      "(1, 0.96289998, 0.97670001)\n",
      "[1.00002, 53.505371, 1.0, 2.8220255, 0.62836474, 0.078432418]\n",
      "(0, 0.97718334, 0.96880001)\n",
      "[1.0000182, 50.858448, 1.0000001, 2.8898001, 0.36982989, 0.023196533]\n",
      "(1, 0.96026665, 0.97409999)\n",
      "[1.0000166, 53.494473, 1.0000001, 2.8843706, 0.5913288, 0.058508299]\n",
      "(0, 0.96886665, 0.96249998)\n",
      "[50.000179, 50.880085, 0.99999994, 2.7797565, 0.36600924, 0.025523931]\n",
      "(1, 0.9691, 0.9777)\n",
      "[1.0000161, 53.464817, 1.0, 2.8027086, 0.56046993, 0.046254572]\n",
      "(0, 0.97746664, 0.96850002)\n",
      "[49.945312, 50.875057, 1.0, 2.6802123, 0.37153935, 0.031189574]\n",
      "(1, 0.95486665, 0.97049999)\n",
      "[1.0000174, 53.505562, 0.9999997, 2.6939344, 0.56556177, 0.053118594]\n",
      "(0, 0.97201669, 0.96450001)\n",
      "[49.974472, 50.884235, 1.0, 2.7990813, 0.36775476, 0.018269742]\n",
      "(1, 0.96684998, 0.97850001)\n",
      "[1.0000169, 53.521572, 1.0000001, 2.8128562, 0.60259563, 0.048026178]\n",
      "(0, 0.96745002, 0.95929998)\n",
      "[1.0000142, 50.893742, 0.9999997, 2.7706735, 0.36303023, 0.017304901]\n",
      "(1, 0.95856667, 0.97229999)\n",
      "[1.0000166, 53.518299, 1.0000005, 2.7763581, 0.58267349, 0.05574888]\n",
      "(0, 0.96658331, 0.9594)\n",
      "[1.0000169, 50.900955, 0.99999994, 2.8199363, 0.38292295, 0.021487147]\n",
      "(1, 0.96534997, 0.9777)\n",
      "[1.0000198, 53.519127, 1.0000002, 2.8448427, 0.62231266, 0.057282262]\n",
      "(0, 0.97158331, 0.96520001)\n",
      "[1.0000142, 50.885605, 1.0, 2.7808688, 0.36702716, 0.029615184]\n",
      "(1, 0.96156669, 0.97539997)\n",
      "[1.0000188, 53.509956, 1.0000001, 2.7942939, 0.60482466, 0.05875013]\n",
      "(0, 0.97216666, 0.9648)\n",
      "[1.0000163, 50.883854, 0.9999997, 2.8466547, 0.3594318, 0.024551963]\n",
      "(1, 0.96378332, 0.97850001)\n",
      "[1.0000143, 53.471432, 1.0, 2.8544452, 0.57157904, 0.054036554]\n",
      "(0, 0.97176665, 0.96329999)\n",
      "[1.0000132, 50.866257, 1.0000001, 2.8640864, 0.35872853, 0.034733884]\n",
      "(1, 0.96301669, 0.97460002)\n",
      "[1.0000147, 53.479336, 1.0000001, 2.8921237, 0.5673995, 0.064091228]\n",
      "(0, 0.96761668, 0.9601)\n",
      "[1.0000162, 50.944897, 1.0, 2.9703588, 0.35505235, 0.0334078]\n",
      "(1, 0.9594667, 0.97500002)\n",
      "[1.0000135, 53.582108, 0.99999988, 2.9955289, 0.57659781, 0.06396661]\n",
      "(0, 0.96688336, 0.95840001)\n",
      "[49.910343, 50.823627, 1.0, 2.7429287, 0.37550095, 0.023924509]\n",
      "(1, 0.96305001, 0.97759998)\n",
      "[1.0000181, 53.475273, 1.0000002, 2.7562165, 0.6027801, 0.059689872]\n",
      "(0, 0.97096664, 0.96170002)\n",
      "[49.987122, 51.023598, 1.0, 2.8698938, 0.36229542, 0.023204727]\n",
      "(1, 0.95964998, 0.97210002)\n",
      "[1.0000144, 53.641632, 0.99999976, 2.8926907, 0.58490872, 0.062171027]\n",
      "(0, 0.96829998, 0.95920002)\n",
      "[49.98243, 50.908035, 0.99999994, 2.9160819, 0.35971338, 0.016034842]\n",
      "(1, 0.96016669, 0.97479999)\n",
      "[1.0000153, 53.546082, 0.99999994, 2.9173012, 0.58591342, 0.050657798]\n",
      "(0, 0.97293335, 0.96600002)\n",
      "[1.0000153, 50.916275, 0.99999976, 2.6384926, 0.37320498, 0.021340176]\n",
      "(1, 0.95021665, 0.96820003)\n",
      "[1.0000172, 53.518982, 1.0, 2.6464961, 0.56827497, 0.05794546]\n",
      "(0, 0.97416669, 0.96759999)\n",
      "[1.0000154, 50.912663, 1.0, 2.824563, 0.34654427, 0.020131141]\n",
      "(1, 0.96724999, 0.97640002)\n",
      "[1.0000166, 53.54187, 0.9999997, 2.8541772, 0.56463641, 0.051846799]\n",
      "(0, 0.97683334, 0.96939999)\n",
      "[1.0000181, 50.829094, 0.99999982, 2.9453177, 0.36336365, 0.022657612]\n",
      "(1, 0.95866668, 0.96990001)\n",
      "[1.0000166, 53.477966, 1.0000002, 2.9552281, 0.59471953, 0.052734748]\n",
      "(0, 0.97031665, 0.96420002)\n",
      "[1.0000162, 50.85844, 1.0000002, 2.7105918, 0.35948235, 0.025067057]\n",
      "(1, 0.96728331, 0.97850001)\n",
      "[1.0000136, 53.429028, 1.0, 2.7396498, 0.60187787, 0.061717246]\n",
      "(0, 0.97108334, 0.96469998)\n",
      "[1.0000138, 50.950661, 1.0, 2.7300141, 0.36362621, 0.018854691]\n",
      "(1, 0.95835, 0.97219998)\n",
      "[1.0000143, 53.548889, 1.0, 2.7258828, 0.58921623, 0.058547691]\n",
      "(0, 0.97348332, 0.96499997)\n",
      "[1.0000151, 50.876026, 1.0, 2.8384378, 0.36411962, 0.029090373]\n",
      "(1, 0.95873332, 0.97149998)\n",
      "[1.0000168, 53.538853, 1.0, 2.842608, 0.58548534, 0.058938589]\n",
      "(0, 0.96569997, 0.96060002)\n",
      "[1.0000174, 50.951996, 1.0, 2.6482937, 0.38034156, 0.030567726]\n",
      "(1, 0.96261668, 0.97549999)\n",
      "[1.0000197, 53.5434, 1.0000004, 2.656605, 0.59660399, 0.051807038]\n",
      "(0, 0.96915001, 0.95999998)\n",
      "[1.0000147, 50.877411, 1.0000001, 2.8855977, 0.35588226, 0.020556688]\n",
      "(1, 0.96485001, 0.98030001)\n",
      "[1.0000184, 53.486538, 1.0, 2.8832166, 0.5827474, 0.056628235]\n",
      "(0, 0.97111666, 0.96240002)\n",
      "[49.950974, 50.898388, 0.99999976, 2.8902209, 0.36267099, 0.02657065]\n",
      "(1, 0.96759999, 0.97890002)\n",
      "[1.0000187, 53.501606, 1.0000001, 2.8868482, 0.56380886, 0.054169957]\n",
      "(0, 0.9655, 0.95779997)\n",
      "[1.0000172, 50.912449, 0.99999994, 2.975903, 0.3757343, 0.024124507]\n",
      "(1, 0.94778335, 0.96859998)\n",
      "[1.0000184, 53.477219, 0.99999982, 2.9857862, 0.63630509, 0.060558025]\n"
     ]
    }
   ],
   "source": [
    "test_accuracies = []\n",
    "train_accuracies = []\n",
    "norms = []\n",
    "\n",
    "for i in range(100):\n",
    "    alignmnist = AlignMNIST()\n",
    "    train_op = tf.contrib.layers.optimize_loss(cost, global_step=gs, learning_rate=lr,optimizer=\"RMSProp\")\n",
    "    max_norm_op = max_norm_constraint(weights, biases)\n",
    "\n",
    "    pred = tf.equal(tf.argmax(mlp, 1), tf.argmax(y, 1)) # Count correct predictions\n",
    "    train_acc_op = tf.reduce_mean(tf.cast(pred, \"float\"))  # Cast boolean to float to average\n",
    "    #tf.scalar_summary(\"training accuracy\", train_acc_op)\n",
    "\n",
    "    test_pred = tf.equal(tf.argmax(mlp_test, 1), tf.argmax(yt, 1)) # Count correct predictions\n",
    "    test_acc_op = tf.reduce_mean(tf.cast(test_pred, \"float\"))  # Cast boolean to float to average\n",
    "    #tf.scalar_summary(\"test 0 accuracy\", test_acc_op)\n",
    "\n",
    "    #writer = tf.train.SummaryWriter(\"./logs/{0}\".format(logfile), sess.graph) # for 0.8\n",
    "    #merged = tf.merge_all_summaries()\n",
    "          \n",
    "    with tf.Session() as sess:\n",
    "        \n",
    "        tf.initialize_all_variables().run()\n",
    "\n",
    "    # create a log writer. run 'tensorboard --logdir=./logs/{logfile}'\n",
    "   \n",
    "        sess.run(max_norm_op)\n",
    "\n",
    "        for epoch in range(2):\n",
    "            trX, trY = alignmnist.next_epoch()\n",
    "\n",
    "            for start, end in zip(range(0, len(trX), batch_size), range(batch_size, len(trX), batch_size)):\n",
    "                sess.run(train_op, feed_dict={x: trX[start:end], y: trY[start:end]})\n",
    "\n",
    "            sess.run(max_norm_op)\n",
    "\n",
    "\n",
    "            trn_acc, tst_acc = sess.run([train_acc_op, test_acc_op], feed_dict={x: trX, y: trY, xt: teX_0, yt: teY_0})\n",
    "            #writer.add_summary(summary, epoch)  \n",
    "\n",
    "\n",
    "            print(epoch, trn_acc, tst_acc)\n",
    "            train_accuracies.append(trn_acc)\n",
    "            test_accuracies.append(tst_acc)\n",
    "\n",
    "            nrms= [np.linalg.norm(weights['h1'].eval()),np.linalg.norm(weights['h2'].eval()), np.linalg.norm(weights['out'].eval()), np.linalg.norm(biases['b1'].eval()),np.linalg.norm(biases['b2'].eval()),np.linalg.norm(biases['out'].eval())]\n",
    "            print nrms\n",
    "            norms.append(nrms)\n",
    "            writer.flush()\n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-17T09:31:16.506974",
     "start_time": "2016-06-17T09:31:16.332666"
    },
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import matplotlib\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-17T09:31:17.162745",
     "start_time": "2016-06-17T09:31:16.546922"
    },
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1590fa9d0>]"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYEAAAD7CAYAAACMlyg3AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nMP8fR/7b5JB5ewQAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1590fa8d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(train_accuracies)\n",
    "plt.plot(test_accuracies)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-17T09:31:17.175044",
     "start_time": "2016-06-17T09:31:17.165161"
    },
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.98269999"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.max(test_accuracies)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Number test errors is"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-17T09:31:21.762744",
     "start_time": "2016-06-17T09:31:21.755734"
    },
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "173.00009727478027"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "10000*(1.0-np.max(test_accuracies))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-17T06:31:18.158336",
     "start_time": "2016-06-17T06:31:18.155316"
    },
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1.000018, 51.163795, 0.9999997, 2.7130308, 0.38692814, 0.026154401]\n",
      "[48.747219, 48.928699, 0.99999994, 2.7886932, 0.27928257, 0.0024301391]\n"
     ]
    }
   ],
   "source": [
    "print norms[0]\n",
    "print norms[-1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-14T16:33:49.528886",
     "start_time": "2016-06-14T16:33:49.525098"
    }
   },
   "source": [
    "#### What is the best way to implement the cutoff ?\n",
    "\n",
    "by the norm of the weights at each layer ?\n",
    "\n",
    "see also:  http://keras.io/constraints/\n",
    "\n",
    "Does this just reflect the loss decreasing ?\n",
    "\n",
    "Can we simply bound the norm?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T22:32:15.292684",
     "start_time": "2016-06-17T05:30:25.992Z"
    },
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "for i in range(6):\n",
    "    plt.plot([nrm[i] for nrm in norms])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-06-16T15:28:16.845184",
     "start_time": "2016-06-16T15:28:16.842225"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 164\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}