Auto_Encoder_Tensorflow_Demo.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Adapted from: https://github.com/lazyprogrammer/machine_learning_examples/blob/master/unsupervised_class3/\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "from sklearn.utils import shuffle"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_mnist(limit=None):\n",
    "  print(\"Reading in and transforming data...\")\n",
    "  df = pd.read_csv('large_files/train.csv')\n",
    "  data = df.values\n",
    "  # np.random.shuffle(data)\n",
    "  X = data[:, 1:] / 255.0 # data is from 0..255\n",
    "  Y = data[:, 0]\n",
    "  X, Y = shuffle(X, Y)\n",
    "  if limit is not None:\n",
    "    X, Y = X[:limit], Y[:limit]\n",
    "  return X, Y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Autoencoder:\n",
    "  def __init__(self, D, M):\n",
    "    # represents a batch of training data\n",
    "    self.X = tf.placeholder(tf.float32, shape=(None, D))\n",
    "\n",
    "    # input -> hidden\n",
    "    self.W = tf.Variable(tf.random_normal(shape=(D, M)) * np.sqrt(2.0 / M))\n",
    "    self.b = tf.Variable(np.zeros(M).astype(np.float32))\n",
    "\n",
    "    # hidden -> output\n",
    "    self.V = tf.Variable(tf.random_normal(shape=(M, D)) * np.sqrt(2.0 / D))\n",
    "    self.c = tf.Variable(np.zeros(D).astype(np.float32))\n",
    "\n",
    "    # construct the reconstruction\n",
    "    self.Z = tf.nn.relu(tf.matmul(self.X, self.W) + self.b)\n",
    "    logits = tf.matmul(self.Z, self.V) + self.c\n",
    "    self.X_hat = tf.nn.sigmoid(logits)\n",
    "\n",
    "    # compute the cost\n",
    "    self.cost = tf.reduce_sum(\n",
    "      tf.nn.sigmoid_cross_entropy_with_logits(\n",
    "        labels=self.X,\n",
    "        logits=logits\n",
    "      )\n",
    "    )\n",
    "\n",
    "    # make the trainer\n",
    "    self.train_op = tf.train.RMSPropOptimizer(learning_rate=0.001).minimize(self.cost)\n",
    "\n",
    "    # set up session and variables for later\n",
    "    self.init_op = tf.global_variables_initializer()\n",
    "    self.sess = tf.InteractiveSession()\n",
    "    self.sess.run(self.init_op)\n",
    "\n",
    "  def fit(self, X, epochs=30, batch_sz=64):\n",
    "    costs = []\n",
    "    n_batches = len(X) // batch_sz\n",
    "    print(\"n_batches:\", n_batches)\n",
    "    for i in range(epochs):\n",
    "      print(\"epoch:\", i)\n",
    "      np.random.shuffle(X)\n",
    "      for j in range(n_batches):\n",
    "        batch = X[j*batch_sz:(j+1)*batch_sz]\n",
    "        _, c, = self.sess.run((self.train_op, self.cost), feed_dict={self.X: batch})\n",
    "        c /= batch_sz # just debugging\n",
    "        costs.append(c)\n",
    "        if j % 100 == 0:\n",
    "          print(\"iter: %d, cost: %.3f\" % (j, c))\n",
    "    plt.plot(costs)\n",
    "    plt.show()\n",
    "\n",
    "  def predict(self, X):\n",
    "    return self.sess.run(self.X_hat, feed_dict={self.X: X})\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Reading in and transforming data...\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "n_batches: 656\nepoch: 0\niter: 0, cost: 562.779\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 112.674\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 84.798\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 79.738\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 73.239\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 68.941\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 64.686\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 1\niter: 0, cost: 62.980\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 67.429\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 67.162\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 60.375\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 57.849\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 58.732\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 65.910\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 2\niter: 0, cost: 58.697\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 60.627\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 60.833\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 55.419\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 58.939\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 56.423\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 63.066\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 3\niter: 0, cost: 56.978\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 55.382\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 56.272\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 52.764\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 55.323\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 58.750\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 54.908\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 4\niter: 0, cost: 53.397\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 57.245\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 52.412\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 53.262\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 54.501\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 53.409\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 53.978\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 5\niter: 0, cost: 54.948\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 53.233\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 51.297\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 51.836\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 51.689\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 54.955\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 50.627\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 6\niter: 0, cost: 55.040\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 49.201\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 54.281\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 55.890\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 54.082\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 52.648\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 54.550\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 7\niter: 0, cost: 52.111\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 52.291\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 52.526\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 55.218\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 53.789\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 51.520\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 53.572\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 8\niter: 0, cost: 55.990\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 53.627\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 53.272\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 51.731\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 52.172\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 54.840\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 52.255\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 9\niter: 0, cost: 52.145\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 50.892\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 51.916\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 50.960\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 47.589\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 49.729\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 50.047\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 10\niter: 0, cost: 53.638\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 52.468\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 53.518\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 55.403\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 56.619\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 53.081\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 53.155\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 11\niter: 0, cost: 51.736\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 50.533\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 53.761\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 54.346\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 53.924\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 53.966\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 51.553\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 12\niter: 0, cost: 52.708\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 53.800\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 55.410\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 49.349\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 52.228\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 49.922\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 49.450\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 13\niter: 0, cost: 51.829\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 54.843\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 54.747\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 52.211\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 50.979\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 53.218\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 51.271\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 14\niter: 0, cost: 50.550\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 51.432\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 51.067\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 50.367\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 51.492\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 47.290\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 51.753\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 15\niter: 0, cost: 51.122\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 51.971\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 50.092\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 49.759\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 51.328\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 49.485\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 51.925\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 16\niter: 0, cost: 51.541\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 52.883\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 51.584\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 52.282\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 50.898\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 52.927\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 53.424\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 17\niter: 0, cost: 51.045\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 50.341\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 51.633\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 51.207\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 52.405\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 51.467\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 50.909\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 18\niter: 0, cost: 51.084\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 55.064\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 49.626\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 51.534\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 52.102\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 54.369\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 49.014\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 19\niter: 0, cost: 51.045\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 50.400\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 54.181\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 50.346\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 50.377\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 49.931\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 51.499\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 20\niter: 0, cost: 52.594\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 53.191\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 50.747\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 47.318\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 51.521\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 49.479\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 52.552\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 21\niter: 0, cost: 51.962\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 49.704\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 53.063\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 51.754\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 52.668\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 52.322\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 51.513\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 22\niter: 0, cost: 49.745\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 50.701\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 51.891\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 51.818\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 50.449\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 52.890\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 49.838\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 23\niter: 0, cost: 52.718\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 50.059\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 52.216\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 49.970\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 48.838\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 50.149\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 50.201\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 24\niter: 0, cost: 52.449\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 49.217\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 51.610\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 50.677\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 50.636\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 51.582\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 47.814\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 25\niter: 0, cost: 50.558\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 50.611\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 50.012\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 49.059\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 52.840\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 52.677\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 50.113\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 26\niter: 0, cost: 50.665\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 52.124\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 48.820\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 55.523\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 50.067\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 51.820\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 49.761\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 27\niter: 0, cost: 52.247\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 50.653\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 49.586\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 48.043\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 48.868\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 50.946\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 51.115\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 28\niter: 0, cost: 51.058\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 49.320\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 49.693\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 51.809\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 48.870\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 48.544\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 49.747\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 29\niter: 0, cost: 50.101\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 100, cost: 47.712\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 200, cost: 54.739\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 300, cost: 53.248\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 400, cost: 50.569\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 500, cost: 53.802\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iter: 600, cost: 50.446\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAADHCAYAAAAJSqg8AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAAF7BJREFUeJzt3X2QHdV55/HvDwkkIRHeZMQIEASLyLAU4CqtCgeMWSMcwOUVrAMGm8TYgFLZiIpTeEEBHOHsigCLA0vIsisChXgxNjbYEBdrVqgSG4xDjLFsw2JbBEtCSJZAQCQZgZD07B/dEw/qczV35r7M7TO/T5VqZp57uvv0zHMf9e1zulsRgZmZ1d9uI90BMzNrDxd0M7NMuKCbmWXCBd3MLBMu6GZmmXBBNzPLhAv6CJF0haS/a3fbJtYVkqa3Y11mdSDpOUknj3Q/ukGeh94eki4ALgXeC2wEvgH8eUS8MZL92pmkAI6IiBdGui/WPpJWAFOA7cBm4NvAvIjYPJL92pmkq4HpEXF+h9Z/J7A6Iq7qxPp7nY/Q20DSpcB1wH8B9gaOBw4FlkjaI9F+bHd7aKPExyJiEnAc8H7gz0e4P0OmguvSMPkX1yJJvwV8EbgkIr4dEe9ExArgHIqifr6kqyV9XdI9kjYCF5Sxewas5w8lrZS0QdIXJK2QNLt87d/aSjqsPG3yaUmrJL0q6coB65kl6fuS3pC0VtItqf9ULF8R8SvgUYrCjqRxkm4o82WdpP8laUJ/e0lzJC2TtFHSv0g6rYxPlfSwpNckvSDp4gHLXC3pfkl3SdpUntaYOeD1yyW9XL72c0mnlOu9AviEpM2Sfly2/UdJCyV9D3gTOHxg/g/Y3sD3y4mSnizz/CVJF0iaC3wKuKxc/9+XbQe+l8ZJuknSmvLfTZLGla+dLGm1pEslrS/fP59p99+nk1zQW/e7wHjgwYHB8qPu/wFOLUNzgK8D+wD3Dmwr6Sjgf1IkYx/FUf5Bg2z3RGAGcArwF5KOLOPbgT8DJgMfKF//z8PYL6spSQcDpwP9p9WuA36HosBPp8itvyjbzgLuovh0uQ9wErCiXO4+YDUwFfh94BpJpwzY1H8EvlIu9zBwS7nOGcA84N9HxF7A7wErIuLbwDXAVyNiUkQcO2BdfwDMBfYCVg6yf9Mo3lt/A7yn3K9lEbGI4r11fbn+jyUWv5LiE/RxwLHALGDg6ZkD+c3770LgbyXtu6v+9BIX9NZNBl6NiG2J19aWrwN8PyK+GRE7ImLLTu1+H/j7iHgiIrZSvNkGG9z4YkRsiYgfAz+mSE4i4ocR8U8Rsa38pPC/gQ8Nb9esZr4paRPwErAeWCBJwMXAn0XEaxGxiaKonlsucyFwR0QsKXPz5Yj4maRDKA4aLo+ItyJiGfB3FIW33xMR8UhEbAfupsxBioOKccBRknaPiBUR8S+D9P3OiHiuzNt3Bmn7KeCxiLiv/ES8oexfMz4F/GVErI+IVyg+XQ/cp3fK19+JiEcoxiNmNLnuEeeC3rpXgckNzov3la9D8SZrZOrA1yPiTWDDINv91YDv3wQmAUj6HUnfkvSr8vTONfzmPxXL25nlEfHJwPso/u7vAfYEflienniDYsD0PeUyhwCpYjsV6P8PoN9K3v3JceccHC9pbDng/jngamC9pK9ImjpI33f1/thZoz43Yyrv/gSwsoz127DTwdm/vbfqwAW9dd8H3gb+08CgpIkUH3uXlqFdHXGvBQ4esOwEYP9h9udW4GcUM1l+i+KcpYa5LquhiPgOcCdwA8UBxRbg30XEPuW/vcvBUygK6XsTq1kD7CdprwGxacDLTfbhyxFxIsU4UlCc9oHG74Od47+m+I+o34EDvm/U512tv9+ask/9ppWxLLigtygi/pXiY9vfSDpN0u6SDgO+RnH+8e4mVvN14GOSfrccwPwiwy/Ce1FMm9ws6X3AHw9zPVZvN1GM3xwD3AbcKOkAAEkHSfq9st3twGfKQcvdytfeFxEvAU8CfyVpvKRjKE7P3Fvd1LtJmiHpw+Vg41sU/6FsL19eBxymwWeyLAPOLd9PMylOS/a7F5gt6RxJYyXtL+m4Aes/fBfrvQ+4StJ7JE2mOL15zy7a14oLehtExPUUR8I3UBTTpyiOIk6JiLebWP454BKKAaa1wCaKc6CDLpvweeCT5TpuA746jHVYzZXnh+8CvgBcTjFA+k/labjHKM8LR8Q/A58BbgT+FfgOvzmCPQ84jOII9hvAgohY0sTmxwHXUnw6+BVwAMX7A4oDHYANkp7ZxTq+QHEU/jrFAc6XB+zbKuAMius+XqMo/v3n72+nOHf/hqRvJtb734CngZ8APwWeKWNZ8IVFPUjSJOANitMmvxzp/phZPfgIvUdI+pikPctz7zdQHD2sGNlemVmduKD3jjkUH23XAEcA54Y/PpnZEPiUi5lZJnyEbmaWiZYKejlN7+flfR7mt6tTZiPNuW11NOxTLpLGAL+gmOu6GvgBcF5E/L9dLOPzO9ZREdHyRVTObetFzeR2K0fos4AXIuLF8v4jX6EY2DOrO+e21VIrBf0g3n3/hdUk7hAoaa6kpyU93cK2zLrJuW211MqDFlKH/5WPneUtLReBP5ZabTi3rZZaOUJfTXHXs34Hk9FNbmxUc25bLbVS0H8AHCHpt8sbSp1LcZN7s7pzblstDfuUS0RskzSP4lFXYyhukv9c23pmNkKc21ZXXb1S1OcZrdPaMW1xOJzb1mmdnrZoZmY9xAXdzCwTLuhmZplwQTczy4QLuplZJlzQzcwy4YJuZpYJF3Qzs0y4oJuZZcIF3cwsEy7oZmaZcEE3M8uEC7qZWSZc0M3MMuGCbmaWCRd0M7NMuKCbmWXCBd3MLBPDfqYogKQVwCZgO7AtIma2o1O94thjj63ELrnkkpbXO23atEps9uzZTS8vVZ9ENZRHCZ511lnJ+EMPPdT0OnKXS27vtlv6mG3WrFmV2M0335xs+/rrr1difX19ybYTJkyoxA455JBKbMyYMcnlh5Lb27Ztq8TOP//8ZNsHH3yw6fXWWUsFvfQfIuLVNqzHrNc4t61WfMrFzCwTrRb0AP6vpB9KmtuODpn1COe21U6rp1xOiIg1kg4Alkj6WUR8d2CD8s3gN4TVjXPbaqelI/SIWFN+XQ98A6iMtETEooiYWddBJRudnNtWRxruSK+kicBuEbGp/H4J8JcR8e1dLFOrYeXly5dXYocffniybaszT4ai1W298MILyfiHP/zhSuzll19uvmM9ICKqv5whyim3x45NfwhftWpVJXbggQcm2+7YsaMSS+VgI6ncbLR8q7n9yiuvJOOpmWVbt25ter29oJncbuWUyxTgG+UfYCzw5V0lvFmNOLetloZd0CPiRaA6Udus5pzbVleetmhmlgkXdDOzTLTjStFs7b333iPdhY6YPn16Mn7ooYdWYnUbFLV3GzduXDK+xx57VGKNBipTtw9oNFCZir/zzjuVWKNL/1PbSg3KQnrAd6+99kq2Tb2XGw2g1pmP0M3MMuGCbmaWCRd0M7NMuKCbmWXCBd3MLBOe5bIL8+bNq8RuvPHGZNvUZcSpy43NumnLli3J+Cc+8YlK7JZbbkm2PfjggyuxiRMnJtumZsqkZqM0evBGapZMoxkxzW4fYPfdd296HXXmI3Qzs0y4oJuZZcIF3cwsEy7oZmaZGPb90Ie1sR69Z/RQTJkyJRlPDT597WtfS7adPXt2S31o9Z7Rjz32WDJ+9tlnV2IbN25svmM9oB33Qx+OuuV2aqCx0b3TzzjjjEps8eLFybaTJk2qxFK5uX379uTyqXjqNgWQfh889NBDybap3N62bVuyba9qJrd9hG5mlgkXdDOzTLigm5llwgXdzCwTgxZ0SXdIWi/p2QGx/SQtkbS8/LpvZ7tp1n7ObctNM5f+3wncAtw1IDYfWBoR10qaX/58efu713vWrVuXjM+fP78SO/XUUzvSh6E8cX3Dhg2V2JVXXplsW7cZLW1wJ6M0txvNMkmZMWNGJdbowRnNanTpf8rbb7+djKfyNfU+hPRDMhq9j7o586/dBv2tRsR3gdd2Cs8B+uctLQbObHO/zDrOuW25Ge459CkRsRag/HpA+7pkNqKc21ZbHb/boqS5wNxOb8es25zb1muGe4S+TlIfQPl1faOGEbEoImZGxMxhbsusm5zbVlvDPUJ/GPg0cG35NX297Shy1llnVWLdHFxptK3UZdOTJ09Otk3dMzr1xPbMjdrcTl22D7BgwYJKrNH9xVODjylDeW80GrwcP358JXb66acn265ataoSazTYmvWgqKT7gO8DMyStlnQhRbKfKmk5cGr5s1mtOLctN4MeoUfEeQ1eOqXNfTHrKue25cZXipqZZcIF3cwsEy7oZmaZ8AMu2iQ1ut+p322rD7ho5I477qjELr744pbX201+wMXw7bnnnsn4a6/tfDFt44dOpB4akbrMv1G+DuW2Fqn1Nppl89RTT1ViH/nIR5Jtf/3rXzfdh27yAy7MzEYRF3Qzs0y4oJuZZcIF3cwsEx4UbZO77767EvvkJz/Z9PJbt26txF588cVk276+vkps7733bnpbQzFv3rxk/NZbb+3I9lrlQdHhGzs2fZ3hddddV4ldeOGFybapgcrUIGNq8BRg4sSJTfdrwoQJTbdN1bnLLrss2fZLX/pSU8t3mwdFzcxGERd0M7NMuKCbmWXCBd3MLBMeFG2T6dOnV2IHHND808vefPPNSmzZsmUtb2v27NmV2OWXp595nLq/9JNPPpls+8EPfjAZH2keFG3OUK7ITN37fMyYMcm2qUHJVNtNmzYll09dgZoaKIX0VcwLFy5Mtk3t7+rVq5NtjzzyyEps8+bNybbd5EFRM7NRxAXdzCwTLuhmZplwQTczy0QzzxS9Q9J6Sc8OiF0t6WVJy8p/Z3S2m2bt59y23Aw6y0XSScBm4K6IOLqMXQ1sjogbhrSxms0EyNUrr7ySjO+///6VWKPZEKlZLk888URrHWuDocxyGc25PZR7iY+0Rjk4adKkSmzNmjVNt92+fXuy7Yc+9KFKrNFsr27OEmzLLJeI+C5QvcO9Wc05ty03rZxDnyfpJ+XH1n3b1iOzkefctloabkG/FXgvcBywFqjenqwkaa6kpyU9PcxtmXWTc9tqa1gFPSLWRcT2iNgB3AbM2kXbRRExMyJmDreTZt3i3LY6S988eBCS+iJibfnjWcCzu2pvI+foo4+uxFKX+MPQBng++tGPVmK9MCjaqtxyeyiX+DdqO9L3Am+0/dTtLsaNG9f0elMDwwBz5sypxBoNivaaQQu6pPuAk4HJklYDC4CTJR0HBLAC+KMO9tGsI5zblptBC3pEnJcI396Bvph1lXPbcuMrRc3MMuGCbmaWCRd0M7NMDGuWi/We1GwWgEcffbQS23PPPTvdHRsBqVkbjWaIDGX2S0o3Z740yu3UzJPUwzgaabQPGzdubHodvcZH6GZmmXBBNzPLhAu6mVkmXNDNzDIx6P3Q27qxmt0zuldddNFFldiCBQuSbadOndrStlauXJmMn3DCCZXY2rVrEy27ayj3Q2+nXsjt1KDoUAY/h1ILGrVtdh1jxoxJxo8//vhKbMmSJcm2EyZMaGpbjfrV6LkAxxxzTCW2fv36ptfbKW25H7qZmdWDC7qZWSZc0M3MMuGCbmaWCRd0M7NM+NL/Xdhnn30qsUaXIbf6cIdDDz20ErvnnnuSbVMzTDo12r5o0aJkvBdmtNi77dixoxJrNJtk332rj0qdPn16su0vfvGLSqzR5fGpWTWpbT3wwAPJ5T/wgQ9UYo32YSi2b99eiS1evDjZNjX7ZaQf8tEsH6GbmWXCBd3MLBMu6GZmmXBBNzPLxKCX/ks6BLgLOBDYASyKiP8haT/gq8BhFA/TPSciXh9kXfUYWSh973vfq8RSgzaQHhRNDSYBTJs2rRKbPXt20/1KDTy1Y9AmNVB19tlnt7zebhrKpf+jIbfHjk3Pe/jRj35Uic2YMSPZ9qWXXqrEHnnkkWTbE088san1jhs3Lrl86vYFjaRyPjX4Cencvvjii5NtN23a1HQfuqldl/5vAy6NiCOB44E/kXQUMB9YGhFHAEvLn83qxLltWRm0oEfE2oh4pvx+E/A8cBAwB+if97MYOLNTnTTrBOe25WZI89AlHQa8H3gKmBIRa6F4Y0g6oMEyc4G5rXXTrLOc25aDpgu6pEnAA8DnImJjs7fljIhFwKJyHT15ntFGN+e25aKpEQhJu1Mk/L0R8WAZXiepr3y9D0jfMNishzm3LSfNzHIRxXnE1yLicwPi/x3YEBHXSpoP7BcRlw2yrlodxaRuar///vsn23Zq5kkntnX//fcn45/97GcrsS1btjTfsR4wxFku2ef2xIkTk/HU5e2NHhiRuqVAKgbNX6bfjgdvbNu2rRK7+eabk22vuuqqSuytt95qug+9oJncbuaUywnAHwA/lbSsjF0BXAvcL+lCYBVQr/ltZs5ty8ygBT0ingAa/c9wSnu7Y9Y9zm3Lja8UNTPLhAu6mVkmfD/0TGzdujUZ//znP1+J3XbbbUNah9VDaqCx0eBlakBw/PjxTa93KJfoD2UANKVRXi5YsKASu+mmm5Jt33777Zb6UBc+Qjczy4QLuplZJlzQzcwy4YJuZpYJF3Qzs0wMeul/WzfWo5dHN3LOOedUYrfffnuybeoS6079bh9//PFK7IILLki2XbFiRUf60KuGcul/O/Vqbje6FP+iiy6qxG644YZk21Rut3rpf6MHUfzyl7+sxE466aRk29StORr1q5t1rlPa9YALMzOrARd0M7NMuKCbmWXCBd3MLBMeFB2ij3/848n4aaed1nTbN954oxJbunRpJbZ8+fLk8tdff/2uujiqeVC0OWPHVu/60dfXl2x7zTXXVGLHHHNMsu2kSZMqsZUrV1ZiCxcuTC6fGvD3LSkKHhQ1MxtFXNDNzDLhgm5mlgkXdDOzTAxa0CUdIukfJD0v6TlJf1rGr5b0sqRl5b8zOt9ds/ZxbltuBp3lIqkP6IuIZyTtBfwQOBM4B9gcEenrhdPrqtVMAKufocxycW5bnTST2808JHotsLb8fpOk54GDWu+e2chybltuhnQOXdJhwPuBp8rQPEk/kXSHpH0bLDNX0tOSnm6pp2Yd5Ny2HDR9YZGkScB3gIUR8aCkKcCrQAD/leKj62cHWYc/llpHDefCIue21UEzud1UQZe0O/At4NGI+OvE64cB34qIowdZj5PeOmqoBd25bXXRlitFVTyy+3bg+YEJXw4o9TsLeHY4nTQbKc5ty00zs1xOBB4Hfgr03z3+CuA84DiKj6UrgD8qB5l2tS4fxVhHDXGWi3PbaqNtp1zaxUlvneabc1mufHMuM7NRxAXdzCwTLuhmZplwQTczy4QLuplZJlzQzcwy4YJuZpYJF3Qzs0wMevvcNnsV6H8E+OTy59x4v0bOoSO47f7crsPvabhy3bc67FdTud3VK0XftWHp6YiYOSIb7yDv1+iW8+8p133Lab98ysXMLBMu6GZmmRjJgr5oBLfdSd6v0S3n31Ou+5bNfo3YOXQzM2svn3IxM8tE1wu6pNMk/VzSC5Lmd3v77VQ+QHi9pGcHxPaTtETS8vJr8gHDvUzSIZL+QdLzkp6T9KdlvPb71km55Lbzun771q+rBV3SGOBvgdOBo4DzJB3VzT602Z3AaTvF5gNLI+IIYGn5c91sAy6NiCOB44E/Kf9OOexbR2SW23fivK6lbh+hzwJeiIgXI2Ir8BVgTpf70DYR8V3gtZ3Cc4DF5feLgTO72qk2iIi1EfFM+f0m4HngIDLYtw7KJred1/Xbt37dLugHAS8N+Hl1GcvJlP7nT5ZfDxjh/rSkfOr9+4GnyGzf2iz33M7qb59rXne7oKeeiedpNj1K0iTgAeBzEbFxpPvT45zbNZFzXne7oK8GDhnw88HAmi73odPWSeoDKL+uH+H+DIuk3SmS/t6IeLAMZ7FvHZJ7bmfxt889r7td0H8AHCHptyXtAZwLPNzlPnTaw8Cny+8/DTw0gn0ZFkkCbgeej4i/HvBS7fetg3LP7dr/7UdDXnf9wiJJZwA3AWOAOyJiYVc70EaS7gNOprhb2zpgAfBN4H5gGrAKODsidh5g6mmSTgQeB34K7CjDV1Ccb6z1vnVSLrntvK7fvvXzlaJmZpnwlaJmZplwQTczy4QLuplZJlzQzcwy4YJuZpYJF3Qzs0y4oJuZZcIF3cwsE/8f8Fk/zrDhy+gAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAADHCAYAAAAJSqg8AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAAFfFJREFUeJzt3X20XfOdx/H3Rx4RJEjSREIqRWNmEVZkddTqIK2iNZhSj5W2VmNRXdXFoPpAp4p2RbG0UxNNliiJdlBNrQ5SraepIjTqIYpyEyGSuVKSjCCS7/yx960re9/cc+95uOf88nmtddc993t+e5/fvvd7vnef328/KCIwM7PWt0Vfd8DMzGrDBd3MLBEu6GZmiXBBNzNLhAu6mVkiXNDNzBLhgt5HJF0g6ae1blvBukLSh2qxLrNWIOkpSQf2dT8aQT4OvTYkfR44GxgPrAJ+CXw9Il7vy35tTFIAu0XE833dF6sdSW3ASGA9sAa4AzgzItb0Zb82Juki4EMRcXKd1n8dsDQivlmP9Tc776HXgKSzge8D/wZsB3wE2AWYL2lgSfv+je2hbSaOiIghwERgH+DrfdyfHlPGdamX/IurkqRtge8AX4mIOyJiXUS0AZ8lK+onS7pI0s2SbpC0Cvh8Hruh03pOkbRY0muSviWpTdLH8+f+3lbSuHzYZKqkJZLaJX2j03omS3pQ0uuSlkn6Udk/FUtXRLwK3ElW2JE0SNL0PF+WS7pG0pYd7SUdKWmhpFWS/irp0Dw+WtI8SSslPS/pS52WuUjSLyRdL2l1PqwxqdPz50l6OX/uL5Km5Ou9ADhO0hpJj+dt75H0PUn/A7wJ7No5/zu9Xuf3ywGS/pDn+UuSPi9pGnAScG6+/l/nbTu/lwZJulLSK/nXlZIG5c8dKGmppLMlrcjfP1+o9d+nnlzQq7c/MBi4tXMw/6j738An8tCRwM3AUODGzm0l7Qn8B1kyjiLby9+pm9c9ANgDmAJ8W9KEPL4e+BqwI/BP+fNn9GK7rEVJGgMcBnQMq30f2J2swH+ILLe+nbedDFxP9ulyKPAxoC1fbi6wFBgNHANcImlKp5f6F+CmfLl5wI/yde4BnAnsFxHbAJ8E2iLiDuAS4OcRMSQi9u60rs8B04BtgMXdbN/OZO+tq4Hh+XYtjIgZZO+tH+TrP6Jk8W+QfYKeCOwNTAY6D898gPfef6cCP5Y0bFP9aSYu6NXbEWiPiHdLnluWPw/wYETcFhEbImLtRu2OAX4dEQ9ExDtkb7buJje+ExFrI+Jx4HGy5CQiHo2IP0bEu/knhf8E/rl3m2Yt5jZJq4GXgBXAhZIEfAn4WkSsjIjVZEX1+HyZU4FZETE/z82XI+IZSWPJdhrOi4i3ImIh8FOywtvhgYj4TUSsB35GnoNkOxWDgD0lDYiItoj4azd9vy4insrzdl03bU8CfhsRc/NPxK/l/avEScC/R8SKiPhfsk/XnbdpXf78uoj4Ddl8xB4VrrvPuaBXrx3YsYtx8VH585C9yboyuvPzEfEm8Fo3r/tqp8dvAkMAJO0u6XZJr+bDO5fw3j8VS9tR+R7xgcCHyf7uw4GtgEfz4YnXySZMh+fLjAXKiu1ooOMfQIfFvP+T48Y5OFhS/3zC/SzgImCFpJskje6m75t6f2ysqz5XYjTv/wSwOI91eG2jnbO/v7dagQt69R4E3gb+tXNQ0tZkH3vvzkOb2uNeBozptOyWwA697M9PgGfIjmTZlmzMUr1cl7WgiLgXuA6YTrZDsRb4h4gYmn9tl0+eQlZIx5es5hVge0nbdIrtDLxcYR/mRMQBZPNIQTbsA12/DzaO/x/ZP6IOH+j0uKs+b2r9HV7J+9Rh5zyWBBf0KkXEG2Qf266WdKikAZLGAf9FNv74swpWczNwhKT98wnM79D7IrwN2WGTayR9GDi9l+ux1nYl2fzNXsC1wBWSRgBI2knSJ/N2M4Ev5JOWW+TPfTgiXgL+AFwqabCkvciGZ24svtT7SdpD0sH5ZONbZP9Q1udPLwfGqfsjWRYCx+fvp0lkw5IdbgQ+LumzkvpL2kHSxE7r33UT650LfFPScEk7kg1v3rCJ9i3FBb0GIuIHZHvC08mK6UNkexFTIuLtCpZ/CvgK2QTTMmA12Rhot8uWOAc4MV/HtcDPe7EOa3H5+PD1wLeA88gmSP+YD8P9lnxcOCIeBr4AXAG8AdzLe3uwJwDjyPZgfwlcGBHzK3j5QcBlZJ8OXgVGkL0/INvRAXhN0mObWMe3yPbC/0a2gzOn07YtAQ4nO+9jJVnx7xi/n0k2dv+6pNtK1nsxsAD4M/AE8FgeS4JPLGpCkoYAr5MNm7zY1/0xs9bgPfQmIekISVvlY+/TyfYe2vq2V2bWSlzQm8eRZB9tXwF2A44Pf3wysx7wkIuZWSK8h25mloiqCnp+mN5f8us8nF+rTpn1Nee2taJeD7lI6gc8S3as61LgEeCEiHh6E8t4fMfqKiKqPonKuW3NqJLcrmYPfTLwfES8kF9/5CayiT2zVufctpZUTUHfifdff2EpJVcIlDRN0gJJC6p4LbNGcm5bS6rmRgtlu/+Fj535JS1ngD+WWstwbltLqmYPfSnZVc86jCGhi9zYZs25bS2pmoL+CLCbpA/mF5Q6nuwi92atzrltLanXQy4R8a6kM8luddWP7CL5T9WsZ2Z9xLltraqhZ4p6nNHqrRaHLfaGc9vqrd6HLZqZWRNxQTczS4QLuplZIlzQzcwS4YJuZpYIF3Qzs0S4oJuZJcIF3cwsES7oZmaJcEE3M0uEC7qZWSJc0M3MEuGCbmaWCBd0M7NEuKCbmSXCBd3MLBEu6GZmiXBBNzNLRK/vKQogqQ1YDawH3o2ISbXolFlfc25bK6qqoOcOioj2GqzHrNk4t62leMjFzCwR1Rb0AO6S9KikabXokFmTcG5by6l2yOWjEfGKpBHAfEnPRMR9nRvkbwa/IazVOLet5SgiarMi6SJgTURM30Sb2ryYWRciQrVep3PbmkElud3rIRdJW0vapuMxcAjwZG/XZ9YsnNv1J6n0y6pTzZDLSOCX+R+hPzAnIu6oSa/M+pZz21pSzYZcKnoxfyy1OqvHkEslnNs909XeeCPrUaup65CLmZk1Fxd0M7NE1OJM0ZYybVrxKLPVq1eXtp07d24h9pnPfKa07e67715Vvy655JJCbMOGDRUv//TTT5fG58yZU4iVbRdAW1tbxa9nfWfYsGGl8b333rsQ23rrrUvbLly4sBCbMGFCaduJEycWYl0NmYwcObIQO+mkkwqxbbfdtnT5119/vRBrby8/WffOO+8sxL773e+Wtl2zZk0hluLwjvfQzcwS4YJuZpYIF3Qzs0S4oJuZJcIF3cwsEcmeWPTMM8+UxnfddddCrKvfwdq1awuxwYMHl7bt37+6A4bKjhqo19/mueeeK41/6lOfKsReeOGFuvShXlI7sWjAgAGF2DXXXFPa9sQTTyzEujoa5Z133inEttxyy9K2W2xR3O+r9jT9npxY1JOjvbo6UmvKlCmF2OLFiytebzPwiUVmZpsRF3Qzs0S4oJuZJcIF3cwsEclOih522GGl8WOPPbZRXeiRek2Klk107rDDDqVtH3744UJs//33r7oPjZTapGjZhOS+++5b2vbiiy8uxIYOHVratmzCv+z0eIB+/foVYiNGjChtu2TJkkLsjTfeKMTWrVtXuvyLL75YiB1yyCGlbSdPnlyIdXXQQtmlDvbbb7/Stj2ZhG0kT4qamW1GXNDNzBLhgm5mlggXdDOzRHRb0CXNkrRC0pOdYttLmi/pufx7+QWazZqYc9tS0+1RLpI+BqwBro+If8xjPwBWRsRlks4HhkXEed2+mO+72HAzZ84sxKZOnVratux08OOPP7607bx586rrWJ305CiXVs3tnpw235NT9MuOqIHyo1y6aluWQz05aqRsvWWvD3D66acXYldccUVp27LfTVe5ffPNN2+qi32mJke5RMR9wMqNwkcCs/PHs4Gjetw7sz7m3LbU9HYMfWRELAPIv5cflGrWepzb1rLqfk9RSdOA4o08zVqcc9uaTW/30JdLGgWQf1/RVcOImBERkyJiUi9fy6yRnNvWsnq7hz4PmApcln//Vc16ZDU1Z86cQqyrSdGBAwcWYtttt13N+9Tkmj63e3JJiJ60Xb9+fWm8J5Oa1V6uouy1ulrnLbfcUohdfvnlpW3L7lcwfvz4Hvau+VVy2OJc4EFgD0lLJZ1KluyfkPQc8In8Z7OW4ty21HS7hx4RJ3TxVPEWIGYtxLltqfGZomZmiXBBNzNLhAu6mVki6n4cuvWtP/3pTxXFAPbZZ596d8eaWE8uE9DIG+N09VqrVq0qxN56663StkOGDCnEym7y0eq8h25mlggXdDOzRLigm5klwgXdzCwRnhRN3JgxYyqKmXWlkROgPZmYLZvoHDRoUGnbsm3wpKiZmTUtF3Qzs0S4oJuZJcIF3cwsEZ4UTVzZjXSHDx/eBz0x652uJkrLcnvAgAGlbcsmRdvb26vrWBPyHrqZWSJc0M3MEuGCbmaWCBd0M7NEVHJP0VmSVkh6slPsIkkvS1qYfx1e326a1Z5z21JTyVEu1wE/Aq7fKH5FREyveY+sV7qa3Z80aVLF61i8eHEh9vDDD/e6Ty3gOjbT3C47cqSRp/j3xFZbbVUanzp1asXrKDui5cEHH+x1n5pVt3voEXEfsLIBfTFrKOe2paaaMfQzJf05/9g6rGY9Mut7zm1rSb0t6D8BxgMTgWXA5V01lDRN0gJJC3r5WmaN5Ny2ltWrgh4RyyNifURsAK4FJm+i7YyImBQRlQ/mmvUR57a1sl6d+i9pVEQsy388GnhyU+2t/q666qrSeNmNn5cuXVra9rDDDivEnn322eo61mI2l9xu1gnQ/v2LJemGG24obTt27NhC7O233y5te9BBBxViK1as6GHvml+3BV3SXOBAYEdJS4ELgQMlTQQCaANOq2MfzerCuW2p6bagR8QJJeGZdeiLWUM5ty01PlPUzCwRLuhmZolwQTczS4RvcNGCTjnllELstNMqn7tbtmxZaXxzO6LFms/JJ59ciH3605+uePm77rqrNF6W2xs2bKi8Yy3Ce+hmZolwQTczS4QLuplZIlzQzcwSoUaeAiypOc83blJ77bVXafzOO+8sxIYPH17a9tJLL60oBvDmm2/2oHfNKSLKbxFfZ87tntl3331L42XXKO/qWv933HFHIXbCCWXnisEbb7zRg941p0py23voZmaJcEE3M0uEC7qZWSJc0M3MEuGCbmaWCB/l0sSWLFlSGh89enTFbQ8++OBCrK2trap+NTMf5dJ8Bg4cWIgtXLiwtO2ECRMKsVWrVpW2HT9+fCHW3t7ew961Dh/lYma2GXFBNzNLhAu6mVkiXNDNzBJRyU2ixwLXAx8ANgAzIuIqSdsDPwfGkd1M97MR8bf6dTVte+yxRyE2bNiw0rZlE6BHH310aduUJ0Cr5dyuPak4b1c2ib/LLruULr9u3bpC7Ljjjittu3Llyh72Ln2V7KG/C5wdEROAjwBflrQncD5wd0TsBtyd/2zWSpzblpRuC3pELIuIx/LHq4FFwE7AkcDsvNls4Kh6ddKsHpzblpoe3YJO0jhgH+AhYGRELIPsjSFpRBfLTAOmVddNs/pyblsKKi7okoYAtwBnRcSqsrGyMhExA5iRr8MnX1jTcW5bKio6ykXSALKEvzEibs3DyyWNyp8fBayoTxfN6se5bSmp5CgXATOBRRHxw05PzQOmApfl339Vlx5uJs4555xCbMsttyxte8899xRijz/+eK27lDzndu919Smm7GYUV199dSE2ePDg0uXvv//+Qux3v/tdadsNGzZsqoubpUqGXD4KfA54QlLHBRguIEv2X0g6FVgCHFufLprVjXPbktJtQY+IB4CuBhWn1LY7Zo3j3LbU+ExRM7NEuKCbmSWiR8ehW22ceuqphdgXv/jFQqyra9XPmjWr5n0y64muJkXLLkFx+OGHF2JdTWheeOGFhVjZ5QCsnPfQzcwS4YJuZpYIF3Qzs0S4oJuZJcIF3cwsET7KpY769etXGt9vv/0qWn727Nml8QULFvS6T2Y9VXZEyxZblO8Llh2lUrb8I488Urp8WW53dbSXFXkP3cwsES7oZmaJcEE3M0uEC7qZWSLUyAmHze2uLmPGjCmNt7W1FWJlE0dDhw4tXX716tVV9StlEVHZ7YZqLOXcLsvN4cOHl7Z96aWXCrGygwN23XXXipf3pGimktz2HrqZWSJc0M3MEuGCbmaWCBd0M7NEdFvQJY2V9HtJiyQ9JemrefwiSS9LWph/FS96bNbEnNuWmkpO/X8XODsiHpO0DfCopPn5c1dExPT6da+1nXvuuX3dBds053YvTZ9e/qsZOHBgIbZ27dpCbNWqVaXL+4iW6lRyk+hlwLL88WpJi4Cd6t0xs3pzbltqejSGLmkcsA/wUB46U9KfJc2SNKyLZaZJWiDJV5SypuXcthRUXNAlDQFuAc6KiFXAT4DxwESyvZzLy5aLiBkRMSkiJtWgv2Y159y2VFRU0CUNIEv4GyPiVoCIWB4R6yNiA3AtMLl+3TSrD+e2paTbMXRl5/3OBBZFxA87xUflY5AARwNP1qeLrWv+/Pml8TPOOKMQu/feewuxt99+u+Z9svc4tytTNlF5++23l7Y95phjCrH29vZCbM2aNdV3zAoqOcrlo8DngCckLcxjFwAnSJoIBNAGnFaXHprVj3PbklLJUS4PAGUXhflN7btj1jjObUuNzxQ1M0uEC7qZWSJc0M3MEuEbXFhSfIMLS5VvcGFmthlxQTczS4QLuplZIlzQzcwSUcmZorXUDizOH++Y/5wab1ff2aUPX7sjt1vh99RbqW5bK2xXRbnd0KNc3vfC0oIUr1Ln7dq8pfx7SnXbUtouD7mYmSXCBd3MLBF9WdBn9OFr15O3a/OW8u8p1W1LZrv6bAzdzMxqy0MuZmaJaHhBl3SopL9Iel7S+Y1+/VrKbyC8QtKTnWLbS5ov6bn8e+kNhpuZpLGSfi9pkaSnJH01j7f8ttVTKrntvG69bevQ0IIuqR/wY+AwYE+yO8Ps2cg+1Nh1wKEbxc4H7o6I3YC7859bzbvA2RExAfgI8OX875TCttVFYrl9Hc7rltToPfTJwPMR8UJEvAPcBBzZ4D7UTETcB6zcKHwkMDt/PBs4qqGdqoGIWBYRj+WPVwOLgJ1IYNvqKJncdl633rZ1aHRB3wl4qdPPS/NYSkZ23GA4/z6ij/tTFUnjgH2Ah0hs22os9dxO6m+fal43uqCXXc/Xh9k0KUlDgFuAsyJiVV/3p8k5t1tEynnd6IK+FBjb6ecxwCsN7kO9LZc0CiD/vqKP+9MrkgaQJf2NEXFrHk5i2+ok9dxO4m+fel43uqA/Auwm6YOSBgLHA/Ma3Id6mwdMzR9PBX7Vh33pFUkCZgKLIuKHnZ5q+W2ro9Rzu+X/9ptDXjf8xCJJhwNXAv2AWRHxvYZ2oIYkzQUOJLta23LgQuA24BfAzsAS4NiI2HiCqalJOgC4H3gC2JCHLyAbb2zpbaunVHLbed1629bBZ4qamSXCZ4qamSXCBd3MLBEu6GZmiXBBNzNLhAu6mVkiXNDNzBLhgm5mlggXdDOzRPw/LZlVfCCkUX8AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "if __name__ == '__main__':\n",
    "  X, Y = get_mnist()\n",
    "\n",
    "  model = Autoencoder(784, 300)\n",
    "  model.fit(X)\n",
    "\n",
    "  # plot reconstruction\n",
    "  counter=0\n",
    "  while counter<10:\n",
    "    counter+=1\n",
    "    i = np.random.choice(len(X))\n",
    "    x = X[i]\n",
    "    im = model.predict([x]).reshape(28, 28)\n",
    "    plt.subplot(1,2,1)\n",
    "    plt.imshow(x.reshape(28, 28), cmap='gray')\n",
    "    plt.title(\"Original\")\n",
    "    plt.subplot(1,2,2)\n",
    "    plt.imshow(im, cmap='gray')\n",
    "    plt.title(\"Reconstruction\")\n",
    "    plt.show()\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "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": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}