KerasMNIST.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Import MNIST data "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "done importing\n"
     ]
    }
   ],
   "source": [
    "from keras.datasets import mnist\n",
    "(X_train, y_train), (X_test, y_test) = mnist.load_data()\n",
    "print(\"done importing\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(60000, 28, 28)\n",
      "(60000,)\n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape)\n",
    "print(y_train.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Show some training images & labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUsAAAEKCAYAAACbs3dXAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nIiIiIiIiIiIiIiIiIiIikt3+P9wwwZILwLZ1AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1147e7310>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "for i in range(9):\n",
    "    plt.subplot(3,3,i+1)\n",
    "    plt.imshow(X_train[i], cmap='gray', interpolation='nearest')\n",
    "    plt.title(\"class={}\".format(y_train[i]))\n",
    "    plt.axis('off')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Reshape and normalize the input data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "('Training matrix shape', (60000, 784))\n",
      "('Testing matrix shape', (10000, 784))\n"
     ]
    }
   ],
   "source": [
    "X_train = X_train.reshape(60000, 784)\n",
    "X_test = X_test.reshape(10000, 784)\n",
    "X_train = X_train.astype('float32')\n",
    "X_test = X_test.astype('float32')\n",
    "X_train /= 255\n",
    "X_test /= 255\n",
    "print(\"Training matrix shape\", X_train.shape)\n",
    "print(\"Testing matrix shape\", X_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Make labels one-hot "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "4\n"
     ]
    }
   ],
   "source": [
    "print(y_train[1])\n",
    "print(y_train[2])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from keras.utils import np_utils\n",
    "\n",
    "Y_train = np_utils.to_categorical(y_train, 10)\n",
    "Y_test = np_utils.to_categorical(y_test, 10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 1.  0.  0.  0.  0.  0.  0.  0.  0.  0.]\n",
      "[ 0.  0.  0.  0.  1.  0.  0.  0.  0.  0.]\n"
     ]
    }
   ],
   "source": [
    "print(Y_train[1])\n",
    "print(Y_train[2])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Build network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers.core import Dense, Dropout, Activation\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Dense(512, input_shape=(784,)))\n",
    "model.add(Activation('relu')) \n",
    "model.add(Dropout(0.2))   \n",
    "\n",
    "model.add(Dense(512))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(0.2))   \n",
    "\n",
    "model.add(Dense(10))\n",
    "model.add(Activation('softmax')) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Compile + specify loss and optimizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model.compile(loss='categorical_crossentropy', optimizer='sgd', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/20\n",
      "60000/60000 [==============================] - 5s - loss: 0.2425 - acc: 0.9301 - val_loss: 0.2009 - val_acc: 0.9416\n",
      "Epoch 2/20\n",
      "60000/60000 [==============================] - 5s - loss: 0.2332 - acc: 0.9329 - val_loss: 0.1925 - val_acc: 0.9434\n",
      "Epoch 3/20\n",
      "60000/60000 [==============================] - 5s - loss: 0.2243 - acc: 0.9352 - val_loss: 0.1845 - val_acc: 0.9470\n",
      "Epoch 4/20\n",
      "60000/60000 [==============================] - 5s - loss: 0.2146 - acc: 0.9383 - val_loss: 0.1779 - val_acc: 0.9487\n",
      "Epoch 5/20\n",
      "60000/60000 [==============================] - 5s - loss: 0.2090 - acc: 0.9393 - val_loss: 0.1704 - val_acc: 0.9506\n",
      "Epoch 6/20\n",
      "60000/60000 [==============================] - 5s - loss: 0.1994 - acc: 0.9425 - val_loss: 0.1652 - val_acc: 0.9528\n",
      "Epoch 7/20\n",
      "60000/60000 [==============================] - 5s - loss: 0.1926 - acc: 0.9448 - val_loss: 0.1596 - val_acc: 0.9527\n",
      "Epoch 8/20\n",
      "60000/60000 [==============================] - 6s - loss: 0.1846 - acc: 0.9465 - val_loss: 0.1538 - val_acc: 0.9546\n",
      "Epoch 9/20\n",
      "60000/60000 [==============================] - 6s - loss: 0.1804 - acc: 0.9483 - val_loss: 0.1498 - val_acc: 0.9558\n",
      "Epoch 10/20\n",
      "60000/60000 [==============================] - 7s - loss: 0.1734 - acc: 0.9492 - val_loss: 0.1451 - val_acc: 0.9576\n",
      "Epoch 11/20\n",
      "60000/60000 [==============================] - 6s - loss: 0.1698 - acc: 0.9506 - val_loss: 0.1412 - val_acc: 0.9581\n",
      "Epoch 12/20\n",
      "60000/60000 [==============================] - 6s - loss: 0.1650 - acc: 0.9520 - val_loss: 0.1366 - val_acc: 0.9598\n",
      "Epoch 13/20\n",
      "60000/60000 [==============================] - 6s - loss: 0.1601 - acc: 0.9538 - val_loss: 0.1332 - val_acc: 0.9602\n",
      "Epoch 14/20\n",
      "60000/60000 [==============================] - 6s - loss: 0.1558 - acc: 0.9551 - val_loss: 0.1298 - val_acc: 0.9620\n",
      "Epoch 15/20\n",
      "60000/60000 [==============================] - 6s - loss: 0.1513 - acc: 0.9560 - val_loss: 0.1274 - val_acc: 0.9616\n",
      "Epoch 16/20\n",
      "60000/60000 [==============================] - 6s - loss: 0.1480 - acc: 0.9570 - val_loss: 0.1241 - val_acc: 0.9631\n",
      "Epoch 17/20\n",
      "60000/60000 [==============================] - 5s - loss: 0.1438 - acc: 0.9582 - val_loss: 0.1213 - val_acc: 0.9635\n",
      "Epoch 18/20\n",
      "60000/60000 [==============================] - 5s - loss: 0.1396 - acc: 0.9589 - val_loss: 0.1192 - val_acc: 0.9648\n",
      "Epoch 19/20\n",
      "60000/60000 [==============================] - 5s - loss: 0.1377 - acc: 0.9608 - val_loss: 0.1172 - val_acc: 0.9652\n",
      "Epoch 20/20\n",
      "60000/60000 [==============================] - 6s - loss: 0.1347 - acc: 0.9605 - val_loss: 0.1136 - val_acc: 0.9664\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x113642c50>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(X_train, Y_train,\n",
    "          batch_size=128, nb_epoch=10,\n",
    "          verbose=1,\n",
    "          validation_data=(X_test, Y_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluate model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 9984/10000 [============================>.] - ETA: 0s('Test score:', 0.20993906867802142)\n",
      "('Test accuracy:', 0.93920000000000003)\n"
     ]
    }
   ],
   "source": [
    "score = model.evaluate(X_test, Y_test, verbose=1)\n",
    "print('Test score:', score[0])\n",
    "print('Test accuracy:', score[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Find correct predictions and mistakes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 9984/10000 [============================>.] - ETA: 0s"
     ]
    }
   ],
   "source": [
    "predicted_classes = model.predict_classes(X_test)\n",
    "\n",
    "import numpy as np\n",
    "\n",
    "correct_indices = np.nonzero(predicted_classes == y_test)[0]\n",
    "incorrect_indices = np.nonzero(predicted_classes != y_test)[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Inspect some correct predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUQAAAEKCAYAAABquCzaAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAAAAAAAAAAAAAAAAAAAAAK5P8BB+Waex1YnwMAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x114e33b10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for i, correct in enumerate(correct_indices[:4]):\n",
    "    plt.subplot(2,2,i+1)\n",
    "    plt.imshow(X_test[correct].reshape(28,28), cmap='gray', interpolation='none')\n",
    "    plt.title(\"predicted={}, true={}\".format(predicted_classes[correct], y_test[correct]))\n",
    "    plt.axis(\"off\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Inspect wrong predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAcMAAAFyCAYAAACJELEOAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1138d6290>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for i, incorrect in enumerate(incorrect_indices[:4]):\n",
    "    plt.subplot(2,2,i+1)\n",
    "    plt.imshow(X_test[incorrect].reshape(28,28), cmap='gray', interpolation='none')\n",
    "    plt.title(\"Predicted={}, True={}\".format(predicted_classes[incorrect], y_test[incorrect]))\n",
    "    plt.axis(\"off\")"
   ]
  }
 ],
 "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": "2.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}