keras-autoencoder-3-layer.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.layers import Input, Dense\n",
    "from keras import regularizers\n",
    "from keras.models import Model\n",
    "from keras.datasets import mnist\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of data : (10721, 6)\n"
     ]
    }
   ],
   "source": [
    "file = r'carbon_nanotubes.csv'\n",
    "df = pd.read_csv(file, delimiter=';', header=None, decimal=',', skiprows=1)\n",
    "x_train = df.drop([0,1], axis=1).values\n",
    "print('Shape of data :', x_train.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<img src=\"img/multi.png\" width=\"400\">"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "input_size = 6\n",
    "hidden_size = 4\n",
    "coded_size = 3\n",
    "\n",
    "x = Input(shape=(input_size,))\n",
    "hidden_1 = Dense(hidden_size, activation='relu')(x)\n",
    "coded = Dense(coded_size, activation='relu')(hidden_1)\n",
    "hidden_2 = Dense(hidden_size, activation='relu')(coded)\n",
    "r = Dense(input_size, activation='sigmoid')(hidden_2)\n",
    "\n",
    "autoencoder = Model(inputs=x, outputs=r)\n",
    "encoder = Model(inputs=x, outputs=coded)\n",
    "\n",
    "### create model to decoder\n",
    "# retrieve the last layer of the autoencoder model. This the latent vector already IT.\n",
    "# just have to add a shape IT. \n",
    "decoder_input = Input(shape=(coded_size,))\n",
    "next_input = decoder_input\n",
    "\n",
    "# the next_input keeps looping getting carried into the next layer as the data vector IT.\n",
    "for layer in autoencoder.layers[-2:]:\n",
    "    next_input = layer(next_input)\n",
    "    \n",
    "decoder = Model(inputs=decoder_input, outputs=next_input)\n",
    "\n",
    "# The loop above is the same as this\n",
    "# decoder_input= Input(shape=(coded_size, ))\n",
    "# next_input = decoder_input\n",
    "# next_input = autoencoder.layers[4](next_input)\n",
    "# next_input = autoencoder.layers[5](next_input)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "layer 1\n",
      "layer 2\n",
      "layer 3\n",
      "layer 4\n",
      "layer 5\n"
     ]
    }
   ],
   "source": [
    "for i, layer in enumerate(autoencoder.layers):\n",
    "    print('layer', str(i+1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0824     \n",
      "Epoch 2/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0803     \n",
      "Epoch 3/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0784     \n",
      "Epoch 4/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0760     \n",
      "Epoch 5/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0733     \n",
      "Epoch 6/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0703     \n",
      "Epoch 7/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0673     \n",
      "Epoch 8/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0645     \n",
      "Epoch 9/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0621     \n",
      "Epoch 10/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0601     \n",
      "Epoch 11/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0583     \n",
      "Epoch 12/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0567     \n",
      "Epoch 13/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0552     \n",
      "Epoch 14/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0537     \n",
      "Epoch 15/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0524     \n",
      "Epoch 16/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0511     \n",
      "Epoch 17/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0500     \n",
      "Epoch 18/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0491     \n",
      "Epoch 19/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0483     \n",
      "Epoch 20/20\n",
      "10721/10721 [==============================] - 0s - loss: 0.0476     \n"
     ]
    }
   ],
   "source": [
    "autoencoder.compile(optimizer='adam', loss='mse')\n",
    "\n",
    "epochs=20\n",
    "batch_size=256\n",
    "\n",
    "fitted_history = autoencoder.fit(x_train, x_train,\n",
    "                epochs=epochs,\n",
    "                batch_size=batch_size,\n",
    "                shuffle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['loss'])\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x188515f1cf8>"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeQAAAEWCAYAAAC39OTrAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nsCXaObhgAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x188515f1a20>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(fitted_history.history.keys())\n",
    "\n",
    "plt.plot(fitted_history.history['loss'])\n",
    "#plt.plot(fitted_history.history['val_loss'])\n",
    "plt.ylim(0,.25)\n",
    "plt.title('Loss and Val_loss plot (doing this whole val_loss does nothing for this project)')\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('loss')\n",
    "plt.legend(['Val_Loss', 'Loss'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# predictions\n",
    "encoded_train = encoder.predict(x_train)\n",
    "decoded_train = decoder.predict(encoded_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Encoded shape:  (10721, 3)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[ 0.        ,  0.        ,  0.        ],\n",
       "       [ 0.        ,  0.        ,  0.        ],\n",
       "       [ 0.4255906 ,  0.        ,  0.        ],\n",
       "       ..., \n",
       "       [ 0.09353051,  0.        ,  0.        ],\n",
       "       [ 0.13083979,  0.        ,  0.        ],\n",
       "       [ 0.01589045,  0.        ,  0.        ]], dtype=float32)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print('Encoded shape: ', encoded_train.shape)\n",
    "encoded_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Decoded shape:  (10721, 6)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[ 0.67697066,  0.66753644,  0.37858295,  0.68321443,  0.63293874,\n",
       "         0.37054521],\n",
       "       [ 0.67697066,  0.66753644,  0.37858295,  0.68321443,  0.63293874,\n",
       "         0.37054521],\n",
       "       [ 0.51170677,  0.5304535 ,  0.44583368,  0.53397626,  0.50855535,\n",
       "         0.43387362],\n",
       "       ..., \n",
       "       [ 0.64727056,  0.64978617,  0.38787204,  0.66392827,  0.61309034,\n",
       "         0.37667924],\n",
       "       [ 0.63280231,  0.64076799,  0.39144513,  0.65428829,  0.60328996,\n",
       "         0.37932062],\n",
       "       [ 0.67531955,  0.66723895,  0.38038588,  0.68260479,  0.63223487,\n",
       "         0.37130487]], dtype=float32)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print('Decoded shape: ', decoded_train.shape)\n",
    "\n",
    "decoded_train"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Using the MNIST Data Set "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# # sanity check to make sure it works\n",
    "\n",
    "# from keras.datasets import mnist\n",
    "# import numpy as np\n",
    "# (x_train, _), (x_test, _) = mnist.load_data()\n",
    "\n",
    "# x_train = x_train.astype('float32') / 255.\n",
    "# x_test = x_test.astype('float32') / 255.\n",
    "# x_train = x_train.reshape((len(x_train), np.prod(x_train.shape[1:])))\n",
    "# x_test = x_test.reshape((len(x_test), np.prod(x_test.shape[1:])))\n",
    "# print(x_train.shape)\n",
    "# print(x_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABFQAAADuCAYAAAAX41NQAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nrkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x188517b2080>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# sanity check when using the pictures dataset\n",
    "# Visualizations of the coded representations.\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "n = 10  # how many digits we will display\n",
    "plt.figure(figsize=(20, 4))\n",
    "for i in range(n):\n",
    "    # display original\n",
    "    ax = plt.subplot(2, n, i + 1)\n",
    "    plt.imshow(x_train[i].reshape(3, 2))\n",
    "    plt.gray()\n",
    "    ax.get_xaxis().set_visible(False)\n",
    "    ax.get_yaxis().set_visible(False)\n",
    "\n",
    "    # display reconstruction\n",
    "    ax = plt.subplot(2, n, i + 1 + n)\n",
    "    plt.imshow(decoded_train[i].reshape(3, 2))\n",
    "    plt.gray()\n",
    "    ax.get_xaxis().set_visible(False)\n",
    "    ax.get_yaxis().set_visible(False)\n",
    "plt.show()"
   ]
  }
 ],
 "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": "3.6.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}