keras.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "import keras\n",
    "\n",
    "import numpy as np\n",
    "import keras.datasets.mnist as mnist\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Keras and MNIST\n",
    "\n",
    "https://keras.io/\n",
    "\n",
    "https://en.wikipedia.org/wiki/MNIST_database"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(60000, 'Number of train samples')\n",
      "(10000, 'Number of test samples')\n",
      "('x_train shape:', (60000, 28, 28))\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11b0632d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11e66fd90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11ddd8b90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nfrNEOfxmifo/Lipt9ZeYlTMAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1217d0590>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11b063a50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# load dataset\n",
    "(x_train, y_train), (x_test, y_test) = mnist.load_data()\n",
    "\n",
    "print(x_train.shape[0], 'Number of train samples')\n",
    "print(x_test.shape[0], 'Number of test samples')\n",
    "print('x_train shape:', x_train.shape)\n",
    "\n",
    "# plot images as gray scale\n",
    "for image_id in range(0, 5):\n",
    "    plt.imshow(x_train[image_id], cmap=plt.get_cmap('gray'))\n",
    "    plt.title(\"The true label is %s\" % str(y_train[image_id]))\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Overfitting and regularization\n",
    "\n",
    "Deeper neural networks face the problem of overfitting."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![Overfitting](https://raw.githubusercontent.com/UCIDataScienceInitiative/LearningWithPython/master/figures/overfitting.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# This makes sure the image has the correct order in the axis for Tensorflow, it would be different for Theano backend\n",
    "x_train = x_train.reshape(x_train.shape[0], 28, 28, 1)\n",
    "x_test = x_test.reshape(x_test.shape[0], 28, 28, 1)\n",
    "\n",
    "# Convert values to floats, originally they are integers\n",
    "x_train = x_train.astype('float32')\n",
    "x_test = x_test.astype('float32')\n",
    "\n",
    "# Scale the inputs\n",
    "x_train /= 255.0  # The image is in grayscale and has values between 0 and 255\n",
    "x_test /= 255.0\n",
    "\n",
    "# Convert values of labels from 0 to 9 to categorical (one_hot encoding)\n",
    "y_train = keras.utils.to_categorical(y_train, 10)\n",
    "y_test = keras.utils.to_categorical(y_test, 10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "((60000, 28, 28, 1), (60000, 10))\n"
     ]
    }
   ],
   "source": [
    "print(x_train.shape, y_train.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A simple way to prevent overfitting is dropout. Dropout randomly removes neurons or connections in the network during training. There are many interpretations of dropout from \"robust\" representation to Bayesian inference.\n",
    "\n",
    "https://www.youtube.com/watch?v=vAVOY8frLlQ\n",
    "\n",
    "http://mlg.eng.cam.ac.uk/yarin/blog_3d801aa532c1ce.html"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![Dropout](https://raw.githubusercontent.com/UCIDataScienceInitiative/LearningWithPython/master/figures/dropout.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Flatten, Dropout, Activation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "model.add(Flatten(input_shape=(28, 28, 1)))  # Images are a 3D matrix, we have to flatten them to be 1D\n",
    "model.add(Dense(200, kernel_initializer='normal', activation='relu'))\n",
    "model.add(Dropout(0.5)) # drop a unit with  50% probability.\n",
    "model.add(Dense(150, kernel_initializer='orthogonal'))\n",
    "model.add(Activation('sigmoid'))\n",
    "model.add(Dense(10, kernel_initializer='normal', activation='softmax')) # last layer, this has a softmax to do the classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "flatten_1 (Flatten)          (None, 784)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 200)               157000    \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 200)               0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 150)               30150     \n",
      "_________________________________________________________________\n",
      "activation_1 (Activation)    (None, 150)               0         \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 10)                1510      \n",
      "=================================================================\n",
      "Total params: 188,660\n",
      "Trainable params: 188,660\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "sgd = keras.optimizers.SGD(lr=0.01, decay=1e-6, momentum=0.9)\n",
    "\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 48000 samples, validate on 12000 samples\n",
      "Epoch 1/10\n",
      "48000/48000 [==============================] - 4s 79us/step - loss: 0.2271 - acc: 0.9357 - val_loss: 0.1407 - val_acc: 0.9577\n",
      "Epoch 2/10\n",
      "48000/48000 [==============================] - 4s 81us/step - loss: 0.1902 - acc: 0.9428 - val_loss: 0.1296 - val_acc: 0.9613\n",
      "Epoch 3/10\n",
      "48000/48000 [==============================] - 5s 97us/step - loss: 0.1737 - acc: 0.9476 - val_loss: 0.1198 - val_acc: 0.9648\n",
      "Epoch 4/10\n",
      "48000/48000 [==============================] - 4s 84us/step - loss: 0.1568 - acc: 0.9530 - val_loss: 0.1116 - val_acc: 0.9683\n",
      "Epoch 5/10\n",
      "48000/48000 [==============================] - 4s 79us/step - loss: 0.1433 - acc: 0.9562 - val_loss: 0.1060 - val_acc: 0.9683\n",
      "Epoch 6/10\n",
      "48000/48000 [==============================] - 4s 78us/step - loss: 0.1335 - acc: 0.9601 - val_loss: 0.1019 - val_acc: 0.9693\n",
      "Epoch 7/10\n",
      "48000/48000 [==============================] - 4s 88us/step - loss: 0.1291 - acc: 0.9601 - val_loss: 0.0984 - val_acc: 0.9708\n",
      "Epoch 8/10\n",
      "48000/48000 [==============================] - 4s 86us/step - loss: 0.1182 - acc: 0.9635 - val_loss: 0.0900 - val_acc: 0.9733\n",
      "Epoch 9/10\n",
      "48000/48000 [==============================] - 4s 82us/step - loss: 0.1086 - acc: 0.9661 - val_loss: 0.0871 - val_acc: 0.9733\n",
      "Epoch 10/10\n",
      "48000/48000 [==============================] - 4s 82us/step - loss: 0.1040 - acc: 0.9687 - val_loss: 0.0865 - val_acc: 0.9748\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(x_train, y_train, validation_split=0.2, epochs=10, batch_size=100, verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "('Keys for history object', ['acc', 'loss', 'val_acc', 'val_loss'])\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEKCAYAAAD9xUlFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQURE0hQKIiKSplAQEZG0/wNz4wNbcJWMugAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x13507fd90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(\"Keys for history object\", history.history.keys())\n",
    "\n",
    "train_loss = history.history['loss']\n",
    "valid_loss = history.history['val_loss']\n",
    "epochs = list(range(1, len(train_loss)+1))\n",
    "\n",
    "plt.plot(epochs, train_loss, label=\"train loss\")\n",
    "plt.plot(epochs, valid_loss, label=\"validation loss\")\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend(loc='upper right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(10000, 10)\n",
      "[7 2 1 0 4 1 4 9 6 9]\n",
      "[7 2 1 0 4 1 4 9 5 9]\n"
     ]
    }
   ],
   "source": [
    "predictions = model.predict(x_test)\n",
    "\n",
    "print(predictions.shape)\n",
    "print(np.argmax(predictions, axis=1)[0:10])\n",
    "print(np.argmax(y_test, axis=1)[0:10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import keras.datasets.cifar10 as cifar10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(50000, 'Number of train samples')\n",
      "(10000, 'Number of test samples')\n",
      "('x_train shape:', (50000, 32, 32, 3))\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12763e250>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nrkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1275d3350>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12757d910>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSSnu/I6TUtz5HSeluPM7Tkr5fxO8jUi4x0DBAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x127776450>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12762c890>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# load dataset\n",
    "(x_train, y_train), (x_test, y_test) = cifar10.load_data()\n",
    "\n",
    "# This is needed to know what each class means\n",
    "label_id_to_class_name = {0:'airplane', 1:'automovile', 2:'bird', 3:'cat', 4:'deer', 5:'dog', \n",
    "                          6:'frog', 7:'horse', 8:'ship', 9:'truck'}\n",
    "\n",
    "print(x_train.shape[0], 'Number of train samples')\n",
    "print(x_test.shape[0], 'Number of test samples')\n",
    "print('x_train shape:', x_train.shape)\n",
    "\n",
    "# plot images\n",
    "for image_id in range(0, 5):\n",
    "    plt.imshow(x_train[image_id])\n",
    "    plt.title(\"The true label is %s\" % label_id_to_class_name[int(y_train[image_id])])\n",
    "    plt.show()\n",
    "    \n",
    "# This makes sure the image has the correct order in the axis for Tensorflow, it would be different for Theano backend\n",
    "x_train = x_train.reshape(x_train.shape[0], 32, 32, 3)\n",
    "x_test = x_test.reshape(x_test.shape[0], 32, 32, 3)\n",
    "\n",
    "# Convert values to floats, originally they are integers\n",
    "x_train = x_train.astype('float32')\n",
    "x_test = x_test.astype('float32')\n",
    "\n",
    "# Convert values of labels from 0 to 9 to categorical (one_hot encoding)\n",
    "y_train = keras.utils.to_categorical(y_train, 10)\n",
    "y_test = keras.utils.to_categorical(y_test, 10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [conda env:tensorflow]",
   "language": "python",
   "name": "conda-env-tensorflow-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}