Non_linear_autoencoder.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "from keras.layers import Dense, Activation, Input\n",
    "from keras.models import Model\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>label</th>\n",
       "      <th>pixel0</th>\n",
       "      <th>pixel1</th>\n",
       "      <th>pixel2</th>\n",
       "      <th>pixel3</th>\n",
       "      <th>pixel4</th>\n",
       "      <th>pixel5</th>\n",
       "      <th>pixel6</th>\n",
       "      <th>pixel7</th>\n",
       "      <th>pixel8</th>\n",
       "      <th>...</th>\n",
       "      <th>pixel774</th>\n",
       "      <th>pixel775</th>\n",
       "      <th>pixel776</th>\n",
       "      <th>pixel777</th>\n",
       "      <th>pixel778</th>\n",
       "      <th>pixel779</th>\n",
       "      <th>pixel780</th>\n",
       "      <th>pixel781</th>\n",
       "      <th>pixel782</th>\n",
       "      <th>pixel783</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 785 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   label  pixel0  pixel1  pixel2  pixel3  pixel4  pixel5  pixel6  pixel7  \\\n",
       "0      1       0       0       0       0       0       0       0       0   \n",
       "1      0       0       0       0       0       0       0       0       0   \n",
       "2      1       0       0       0       0       0       0       0       0   \n",
       "3      4       0       0       0       0       0       0       0       0   \n",
       "4      0       0       0       0       0       0       0       0       0   \n",
       "\n",
       "   pixel8    ...     pixel774  pixel775  pixel776  pixel777  pixel778  \\\n",
       "0       0    ...            0         0         0         0         0   \n",
       "1       0    ...            0         0         0         0         0   \n",
       "2       0    ...            0         0         0         0         0   \n",
       "3       0    ...            0         0         0         0         0   \n",
       "4       0    ...            0         0         0         0         0   \n",
       "\n",
       "   pixel779  pixel780  pixel781  pixel782  pixel783  \n",
       "0         0         0         0         0         0  \n",
       "1         0         0         0         0         0  \n",
       "2         0         0         0         0         0  \n",
       "3         0         0         0         0         0  \n",
       "4         0         0         0         0         0  \n",
       "\n",
       "[5 rows x 785 columns]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ds = pd.read_csv('/home/siddharth/ML/Data_set/train.csv')\n",
    "ds.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(42000, 785)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ds.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(8000, 784)\n"
     ]
    }
   ],
   "source": [
    "data = ds.values[:8000, 1:]\n",
    "print(data.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         (None, 784)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 100)               78500     \n",
      "_________________________________________________________________\n",
      "activation_1 (Activation)    (None, 100)               0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 784)               79184     \n",
      "_________________________________________________________________\n",
      "activation_2 (Activation)    (None, 784)               0         \n",
      "=================================================================\n",
      "Total params: 157,684\n",
      "Trainable params: 157,684\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "inp = Input(shape=(784,))\n",
    "h1 = Dense(100)##first hidden layer\n",
    "a1 = Activation('sigmoid')\n",
    "h2 = Dense(784)##second hidden layer(here it is the output layer)\n",
    "a2 = Activation('sigmoid')\n",
    "out = a2(h2(a1(h1(inp))))\n",
    "\n",
    "model = Model(inputs=[inp], outputs=[out])\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(loss='MSE', optimizer='adam', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(6400, 784) (6400, 784)\n"
     ]
    }
   ],
   "source": [
    "split = int(0.8*data.shape[0])\n",
    "X_train = data[:split]/255.0\n",
    "Y_train = data[:split]/255.0\n",
    "\n",
    "X_test = data[split:]/255.0\n",
    "Y_test = data[split:]/255.0\n",
    "\n",
    "print(X_train.shape, Y_train.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 6400 samples, validate on 1600 samples\n",
      "Epoch 1/60\n",
      "6400/6400 [==============================] - 3s 482us/step - loss: 0.0620 - acc: 0.0133 - val_loss: 0.0461 - val_acc: 0.0169\n",
      "Epoch 2/60\n",
      "6400/6400 [==============================] - 2s 381us/step - loss: 0.0389 - acc: 0.0142 - val_loss: 0.0331 - val_acc: 0.0119\n",
      "Epoch 3/60\n",
      "6400/6400 [==============================] - 2s 380us/step - loss: 0.0295 - acc: 0.0109 - val_loss: 0.0263 - val_acc: 0.0125\n",
      "Epoch 4/60\n",
      "6400/6400 [==============================] - 2s 374us/step - loss: 0.0240 - acc: 0.0116 - val_loss: 0.0220 - val_acc: 0.0119\n",
      "Epoch 5/60\n",
      "6400/6400 [==============================] - 2s 376us/step - loss: 0.0202 - acc: 0.0103 - val_loss: 0.0189 - val_acc: 0.0131\n",
      "Epoch 6/60\n",
      "6400/6400 [==============================] - 2s 378us/step - loss: 0.0174 - acc: 0.0081 - val_loss: 0.0165 - val_acc: 0.0150\n",
      "Epoch 7/60\n",
      "6400/6400 [==============================] - 3s 394us/step - loss: 0.0152 - acc: 0.0089 - val_loss: 0.0147 - val_acc: 0.0131\n",
      "Epoch 8/60\n",
      "6400/6400 [==============================] - 3s 423us/step - loss: 0.0134 - acc: 0.0111 - val_loss: 0.0132 - val_acc: 0.0125\n",
      "Epoch 9/60\n",
      "6400/6400 [==============================] - 3s 426us/step - loss: 0.0121 - acc: 0.0102 - val_loss: 0.0120 - val_acc: 0.0144\n",
      "Epoch 10/60\n",
      "6400/6400 [==============================] - 3s 395us/step - loss: 0.0109 - acc: 0.0105 - val_loss: 0.0110 - val_acc: 0.0125\n",
      "Epoch 11/60\n",
      "6400/6400 [==============================] - 3s 402us/step - loss: 0.0100 - acc: 0.0103 - val_loss: 0.0102 - val_acc: 0.0131\n",
      "Epoch 12/60\n",
      "6400/6400 [==============================] - 3s 404us/step - loss: 0.0091 - acc: 0.0100 - val_loss: 0.0095 - val_acc: 0.0150\n",
      "Epoch 13/60\n",
      "6400/6400 [==============================] - 3s 395us/step - loss: 0.0084 - acc: 0.0114 - val_loss: 0.0088 - val_acc: 0.0156\n",
      "Epoch 14/60\n",
      "6400/6400 [==============================] - 3s 400us/step - loss: 0.0078 - acc: 0.0127 - val_loss: 0.0083 - val_acc: 0.0131\n",
      "Epoch 15/60\n",
      "6400/6400 [==============================] - 3s 409us/step - loss: 0.0073 - acc: 0.0120 - val_loss: 0.0078 - val_acc: 0.0144\n",
      "Epoch 16/60\n",
      "6400/6400 [==============================] - 3s 394us/step - loss: 0.0068 - acc: 0.0125 - val_loss: 0.0074 - val_acc: 0.0138\n",
      "Epoch 17/60\n",
      "6400/6400 [==============================] - 2s 380us/step - loss: 0.0064 - acc: 0.0131 - val_loss: 0.0070 - val_acc: 0.0138\n",
      "Epoch 18/60\n",
      "6400/6400 [==============================] - 2s 348us/step - loss: 0.0060 - acc: 0.0117 - val_loss: 0.0067 - val_acc: 0.0144\n",
      "Epoch 19/60\n",
      "6400/6400 [==============================] - 2s 357us/step - loss: 0.0057 - acc: 0.0106 - val_loss: 0.0064 - val_acc: 0.0169\n",
      "Epoch 20/60\n",
      "6400/6400 [==============================] - 2s 389us/step - loss: 0.0054 - acc: 0.0133 - val_loss: 0.0061 - val_acc: 0.0138\n",
      "Epoch 21/60\n",
      "6400/6400 [==============================] - 2s 383us/step - loss: 0.0051 - acc: 0.0128 - val_loss: 0.0059 - val_acc: 0.0156\n",
      "Epoch 22/60\n",
      "6400/6400 [==============================] - 3s 391us/step - loss: 0.0049 - acc: 0.0122 - val_loss: 0.0058 - val_acc: 0.0144\n",
      "Epoch 23/60\n",
      "6400/6400 [==============================] - 2s 386us/step - loss: 0.0047 - acc: 0.0133 - val_loss: 0.0056 - val_acc: 0.0119\n",
      "Epoch 24/60\n",
      "6400/6400 [==============================] - 3s 402us/step - loss: 0.0045 - acc: 0.0130 - val_loss: 0.0054 - val_acc: 0.0150\n",
      "Epoch 25/60\n",
      "6400/6400 [==============================] - 3s 427us/step - loss: 0.0043 - acc: 0.0123 - val_loss: 0.0052 - val_acc: 0.0119\n",
      "Epoch 26/60\n",
      "6400/6400 [==============================] - 2s 382us/step - loss: 0.0042 - acc: 0.0123 - val_loss: 0.0051 - val_acc: 0.0106\n",
      "Epoch 27/60\n",
      "6400/6400 [==============================] - 2s 327us/step - loss: 0.0040 - acc: 0.0123 - val_loss: 0.0050 - val_acc: 0.0144\n",
      "Epoch 28/60\n",
      "6400/6400 [==============================] - 2s 389us/step - loss: 0.0039 - acc: 0.0131 - val_loss: 0.0049 - val_acc: 0.0125\n",
      "Epoch 29/60\n",
      "6400/6400 [==============================] - 2s 379us/step - loss: 0.0038 - acc: 0.0114 - val_loss: 0.0048 - val_acc: 0.0119\n",
      "Epoch 30/60\n",
      "6400/6400 [==============================] - 3s 392us/step - loss: 0.0037 - acc: 0.0128 - val_loss: 0.0047 - val_acc: 0.0094\n",
      "Epoch 31/60\n",
      "6400/6400 [==============================] - 3s 411us/step - loss: 0.0036 - acc: 0.0113 - val_loss: 0.0046 - val_acc: 0.0119\n",
      "Epoch 32/60\n",
      "6400/6400 [==============================] - 3s 420us/step - loss: 0.0035 - acc: 0.0127 - val_loss: 0.0046 - val_acc: 0.0112\n",
      "Epoch 33/60\n",
      "6400/6400 [==============================] - 3s 397us/step - loss: 0.0034 - acc: 0.0116 - val_loss: 0.0045 - val_acc: 0.0088\n",
      "Epoch 34/60\n",
      "6400/6400 [==============================] - 2s 335us/step - loss: 0.0034 - acc: 0.0119 - val_loss: 0.0045 - val_acc: 0.0131\n",
      "Epoch 35/60\n",
      "6400/6400 [==============================] - 2s 388us/step - loss: 0.0033 - acc: 0.0123 - val_loss: 0.0044 - val_acc: 0.0144\n",
      "Epoch 36/60\n",
      "6400/6400 [==============================] - 2s 391us/step - loss: 0.0032 - acc: 0.0125 - val_loss: 0.0044 - val_acc: 0.0119\n",
      "Epoch 37/60\n",
      "6400/6400 [==============================] - 3s 397us/step - loss: 0.0032 - acc: 0.0125 - val_loss: 0.0042 - val_acc: 0.0125\n",
      "Epoch 38/60\n",
      "6400/6400 [==============================] - 2s 375us/step - loss: 0.0031 - acc: 0.0144 - val_loss: 0.0042 - val_acc: 0.0106\n",
      "Epoch 39/60\n",
      "6400/6400 [==============================] - 2s 386us/step - loss: 0.0031 - acc: 0.0119 - val_loss: 0.0041 - val_acc: 0.0100\n",
      "Epoch 40/60\n",
      "6400/6400 [==============================] - 3s 405us/step - loss: 0.0030 - acc: 0.0133 - val_loss: 0.0041 - val_acc: 0.0150\n",
      "Epoch 41/60\n",
      "6400/6400 [==============================] - 3s 413us/step - loss: 0.0030 - acc: 0.0131 - val_loss: 0.0041 - val_acc: 0.0138\n",
      "Epoch 42/60\n",
      "6400/6400 [==============================] - 3s 393us/step - loss: 0.0029 - acc: 0.0131 - val_loss: 0.0041 - val_acc: 0.0163\n",
      "Epoch 43/60\n",
      "6400/6400 [==============================] - 2s 390us/step - loss: 0.0029 - acc: 0.0144 - val_loss: 0.0040 - val_acc: 0.0138\n",
      "Epoch 44/60\n",
      "6400/6400 [==============================] - 3s 391us/step - loss: 0.0029 - acc: 0.0123 - val_loss: 0.0040 - val_acc: 0.0150\n",
      "Epoch 45/60\n",
      "6400/6400 [==============================] - 3s 452us/step - loss: 0.0028 - acc: 0.0120 - val_loss: 0.0040 - val_acc: 0.0150\n",
      "Epoch 46/60\n",
      "6400/6400 [==============================] - 3s 426us/step - loss: 0.0028 - acc: 0.0130 - val_loss: 0.0040 - val_acc: 0.0125\n",
      "Epoch 47/60\n",
      "6400/6400 [==============================] - 3s 489us/step - loss: 0.0028 - acc: 0.0138 - val_loss: 0.0039 - val_acc: 0.0138\n",
      "Epoch 48/60\n",
      "6400/6400 [==============================] - 3s 439us/step - loss: 0.0027 - acc: 0.0116 - val_loss: 0.0039 - val_acc: 0.0150\n",
      "Epoch 49/60\n",
      "6400/6400 [==============================] - 3s 421us/step - loss: 0.0027 - acc: 0.0128 - val_loss: 0.0039 - val_acc: 0.0150\n",
      "Epoch 50/60\n",
      "6400/6400 [==============================] - 3s 410us/step - loss: 0.0027 - acc: 0.0141 - val_loss: 0.0038 - val_acc: 0.0112\n",
      "Epoch 51/60\n",
      "6400/6400 [==============================] - 3s 417us/step - loss: 0.0027 - acc: 0.0125 - val_loss: 0.0038 - val_acc: 0.0125\n",
      "Epoch 52/60\n",
      "6400/6400 [==============================] - 3s 404us/step - loss: 0.0026 - acc: 0.0128 - val_loss: 0.0039 - val_acc: 0.0156\n",
      "Epoch 53/60\n",
      "6400/6400 [==============================] - 3s 402us/step - loss: 0.0026 - acc: 0.0116 - val_loss: 0.0038 - val_acc: 0.0112\n",
      "Epoch 54/60\n",
      "6400/6400 [==============================] - 3s 393us/step - loss: 0.0026 - acc: 0.0113 - val_loss: 0.0038 - val_acc: 0.0156\n",
      "Epoch 55/60\n",
      "6400/6400 [==============================] - 3s 393us/step - loss: 0.0026 - acc: 0.0130 - val_loss: 0.0038 - val_acc: 0.0131\n",
      "Epoch 56/60\n",
      "6400/6400 [==============================] - 3s 399us/step - loss: 0.0025 - acc: 0.0123 - val_loss: 0.0037 - val_acc: 0.0138\n",
      "Epoch 57/60\n",
      "6400/6400 [==============================] - 2s 386us/step - loss: 0.0025 - acc: 0.0138 - val_loss: 0.0037 - val_acc: 0.0131\n",
      "Epoch 58/60\n",
      "6400/6400 [==============================] - 2s 391us/step - loss: 0.0025 - acc: 0.0119 - val_loss: 0.0037 - val_acc: 0.0144\n",
      "Epoch 59/60\n",
      "6400/6400 [==============================] - 2s 386us/step - loss: 0.0025 - acc: 0.0128 - val_loss: 0.0037 - val_acc: 0.0144\n",
      "Epoch 60/60\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6400/6400 [==============================] - 2s 377us/step - loss: 0.0025 - acc: 0.0120 - val_loss: 0.0037 - val_acc: 0.0131\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7ff3a41d1eb8>"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(X_train, Y_train, batch_size=8, epochs=60, validation_data=[X_test, Y_test])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "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 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sample = X_train[78]\n",
    "plt.imshow(sample.reshape((28,28)), cmap='gray')\n",
    "plt.show()\n",
    "\n",
    "model.predict(X_train)\n",
    "plt.imshow(X_train[78].reshape((28,28)), cmap='gray')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_2 (InputLayer)         (None, 784)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 100)               78500     \n",
      "_________________________________________________________________\n",
      "activation_1 (Activation)    (None, 100)               0         \n",
      "=================================================================\n",
      "Total params: 78,500\n",
      "Trainable params: 78,500\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "##Encoder layer\n",
    "inp = Input(shape=(784,))\n",
    "enc_out = a1(h1(inp))\n",
    "encoder = Model(inputs=[inp], outputs=[enc_out])\n",
    "\n",
    "encoder.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_3 (InputLayer)         (None, 100)               0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 784)               79184     \n",
      "_________________________________________________________________\n",
      "activation_2 (Activation)    (None, 784)               0         \n",
      "=================================================================\n",
      "Total params: 79,184\n",
      "Trainable params: 79,184\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "##decoder layer\n",
    "dec_inp = Input(shape=(100,))\n",
    "dec_out = a2(h2(dec_inp))\n",
    "decoder = Model(inputs=[dec_inp], outputs=[dec_out])\n",
    "\n",
    "decoder.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1000, 784)\n",
      "(1000, 100)\n"
     ]
    }
   ],
   "source": [
    "img_batch = ds.values[20000:21000, 1:]/255.0\n",
    "print(img_batch.shape)\n",
    "\n",
    "ex = encoder.predict(img_batch)\n",
    "print(ex.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1000, 784)\n"
     ]
    }
   ],
   "source": [
    "dx = decoder.predict(ex)\n",
    "print(dx.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "20692\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 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "n = np.random.randint(20000,21000)\n",
    "print(n)\n",
    "\n",
    "plt.imshow(ds.values[n,1:].reshape((28,28)), cmap='gray')\n",
    "plt.show()\n",
    "           \n",
    "plt.imshow(dx[n-20000].reshape((28,28)), cmap='gray')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}