Demo.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "\n",
    "import math\n",
    "import random\n",
    "\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from IPython.display import SVG"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers.core import Dense, Dropout\n",
    "from keras.optimizers import SGD, Adam\n",
    "from keras.utils import np_utils\n",
    "from keras.utils.visualize_util import model_to_dot\n",
    "from keras.datasets import mnist"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Number of epoch\n",
    "NP_EPOCH = 20"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(60000, 28, 28)\n",
      "(10000, 28, 28)\n"
     ]
    }
   ],
   "source": [
    "# Load data from keras\n",
    "(X_train, y_train), (X_test, y_test) = mnist.load_data()\n",
    "\n",
    "print(X_train.shape)\n",
    "print(X_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWUAAAEYCAYAAACEOeekAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nUaEsIpIgKpRFRBJEhbKISIL8fxIsfgecIQ+HAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11c600e48>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for i in range(9):\n",
    "    plt.subplot(3,3,i+1)\n",
    "    plt.imshow(X_train[i], cmap='gray', interpolation='none')\n",
    "    plt.title(\"Class {}\".format(y_train[i]))\n",
    "plt.tight_layout()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 0.  0.  0.  0.  0.  1.  0.  0.  0.  0.]\n"
     ]
    }
   ],
   "source": [
    "X_train = X_train.reshape(60000, 784)\n",
    "X_test = X_test.reshape(10000, 784)\n",
    "\n",
    "X_train = X_train.astype('float32')\n",
    "X_test = X_test.astype('float32')\n",
    "\n",
    "X_train /= 255\n",
    "X_test /= 255\n",
    "\n",
    "Y_train = np_utils.to_categorical(y_train, 10)\n",
    "Y_test = np_utils.to_categorical(y_test, 10)\n",
    "\n",
    "print(Y_train[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [],
   "source": [
    "# Take only 10000 out of 60000\n",
    "\n",
    "X_train = X_train[0:10000]\n",
    "X_test = X_test[0:1000]\n",
    "Y_train = Y_train[0:10000]\n",
    "Y_test = Y_test[0:1000]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Table of Content\n",
    "- Basic Model\n",
    "- With Cross Entropy\n",
    "- Deeper Model\n",
    "- Deeper Model with ReLU\n",
    "- Adam (Adaptive Learning Rate + Momentum)\n",
    "- Dropout"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Basic Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "\n",
    "# input layer\n",
    "model.add(Dense(input_dim=28*28, output_dim=500, activation='sigmoid'))\n",
    "\n",
    "# hidden layer\n",
    "model.add(Dense(output_dim=500, activation='sigmoid'))\n",
    "\n",
    "# output layer\n",
    "model.add(Dense(output_dim=10, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/svg+xml": [
       "<svg height=\"296pt\" viewBox=\"0.00 0.00 225.51 296.00\" width=\"226pt\" xmlns=\"http://www.w3.org/2000/svg\" xmlns:xlink=\"http://www.w3.org/1999/xlink\">\n",
       "<g class=\"graph\" id=\"graph0\" transform=\"scale(1 1) rotate(0) translate(4 292)\">\n",
       "<title>G</title>\n",
       "<polygon fill=\"#ffffff\" points=\"-4,4 -4,-292 221.5107,-292 221.5107,4 -4,4\" stroke=\"transparent\"/>\n",
       "<!-- 4901034864 -->\n",
       "<g class=\"node\" id=\"node1\">\n",
       "<title>4901034864</title>\n",
       "<polygon fill=\"none\" points=\"0,-243.5 0,-287.5 217.5107,-287.5 217.5107,-243.5 0,-243.5\" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"39.0967\" y=\"-261.3\">InputLayer</text>\n",
       "<polyline fill=\"none\" points=\"78.1934,-243.5 78.1934,-287.5 \" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"106.0278\" y=\"-272.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"78.1934,-265.5 133.8623,-265.5 \" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"106.0278\" y=\"-250.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"133.8623,-243.5 133.8623,-287.5 \" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"175.6865\" y=\"-272.3\">(None, 784)</text>\n",
       "<polyline fill=\"none\" points=\"133.8623,-265.5 217.5107,-265.5 \" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"175.6865\" y=\"-250.3\">(None, 784)</text>\n",
       "</g>\n",
       "<!-- 4901035144 -->\n",
       "<g class=\"node\" id=\"node2\">\n",
       "<title>4901035144</title>\n",
       "<polygon fill=\"none\" points=\"13.6035,-162.5 13.6035,-206.5 203.9072,-206.5 203.9072,-162.5 13.6035,-162.5\" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"39.0967\" y=\"-180.3\">Dense</text>\n",
       "<polyline fill=\"none\" points=\"64.5898,-162.5 64.5898,-206.5 \" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"92.4243\" y=\"-191.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"64.5898,-184.5 120.2588,-184.5 \" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"92.4243\" y=\"-169.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"120.2588,-162.5 120.2588,-206.5 \" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"162.083\" y=\"-191.3\">(None, 784)</text>\n",
       "<polyline fill=\"none\" points=\"120.2588,-184.5 203.9072,-184.5 \" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"162.083\" y=\"-169.3\">(None, 500)</text>\n",
       "</g>\n",
       "<!-- 4901034864&#45;&gt;4901035144 -->\n",
       "<g class=\"edge\" id=\"edge1\">\n",
       "<title>4901034864-&gt;4901035144</title>\n",
       "<path d=\"M108.7554,-243.3664C108.7554,-235.1516 108.7554,-225.6579 108.7554,-216.7252\" fill=\"none\" stroke=\"#000000\"/>\n",
       "<polygon fill=\"#000000\" points=\"112.2555,-216.6068 108.7554,-206.6068 105.2555,-216.6069 112.2555,-216.6068\" stroke=\"#000000\"/>\n",
       "</g>\n",
       "<!-- 4901035760 -->\n",
       "<g class=\"node\" id=\"node3\">\n",
       "<title>4901035760</title>\n",
       "<polygon fill=\"none\" points=\"13.6035,-81.5 13.6035,-125.5 203.9072,-125.5 203.9072,-81.5 13.6035,-81.5\" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"39.0967\" y=\"-99.3\">Dense</text>\n",
       "<polyline fill=\"none\" points=\"64.5898,-81.5 64.5898,-125.5 \" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"92.4243\" y=\"-110.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"64.5898,-103.5 120.2588,-103.5 \" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"92.4243\" y=\"-88.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"120.2588,-81.5 120.2588,-125.5 \" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"162.083\" y=\"-110.3\">(None, 500)</text>\n",
       "<polyline fill=\"none\" points=\"120.2588,-103.5 203.9072,-103.5 \" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"162.083\" y=\"-88.3\">(None, 500)</text>\n",
       "</g>\n",
       "<!-- 4901035144&#45;&gt;4901035760 -->\n",
       "<g class=\"edge\" id=\"edge2\">\n",
       "<title>4901035144-&gt;4901035760</title>\n",
       "<path d=\"M108.7554,-162.3664C108.7554,-154.1516 108.7554,-144.6579 108.7554,-135.7252\" fill=\"none\" stroke=\"#000000\"/>\n",
       "<polygon fill=\"#000000\" points=\"112.2555,-135.6068 108.7554,-125.6068 105.2555,-135.6069 112.2555,-135.6068\" stroke=\"#000000\"/>\n",
       "</g>\n",
       "<!-- 4900111920 -->\n",
       "<g class=\"node\" id=\"node4\">\n",
       "<title>4900111920</title>\n",
       "<polygon fill=\"none\" points=\"13.6035,-.5 13.6035,-44.5 203.9072,-44.5 203.9072,-.5 13.6035,-.5\" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"39.0967\" y=\"-18.3\">Dense</text>\n",
       "<polyline fill=\"none\" points=\"64.5898,-.5 64.5898,-44.5 \" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"92.4243\" y=\"-29.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"64.5898,-22.5 120.2588,-22.5 \" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"92.4243\" y=\"-7.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"120.2588,-.5 120.2588,-44.5 \" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"162.083\" y=\"-29.3\">(None, 500)</text>\n",
       "<polyline fill=\"none\" points=\"120.2588,-22.5 203.9072,-22.5 \" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"162.083\" y=\"-7.3\">(None, 10)</text>\n",
       "</g>\n",
       "<!-- 4901035760&#45;&gt;4900111920 -->\n",
       "<g class=\"edge\" id=\"edge3\">\n",
       "<title>4901035760-&gt;4900111920</title>\n",
       "<path d=\"M108.7554,-81.3664C108.7554,-73.1516 108.7554,-63.6579 108.7554,-54.7252\" fill=\"none\" stroke=\"#000000\"/>\n",
       "<polygon fill=\"#000000\" points=\"112.2555,-54.6068 108.7554,-44.6068 105.2555,-54.6069 112.2555,-54.6068\" stroke=\"#000000\"/>\n",
       "</g>\n",
       "</g>\n",
       "</svg>"
      ],
      "text/plain": [
       "<IPython.core.display.SVG object>"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "SVG(model_to_dot(model,\n",
    "                 show_shapes=True,\n",
    "                 show_layer_names=False).create(prog='dot', format='svg'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [],
   "source": [
    "model.compile(loss='mse',\n",
    "              optimizer=SGD(lr=0.1),\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [],
   "source": [
    "def train_and_show_result(model):\n",
    "    training_history = model.fit(X_train, Y_train,\n",
    "                                 batch_size=100, nb_epoch=NP_EPOCH,\n",
    "                                 verbose=2)\n",
    "\n",
    "    score = model.evaluate(X_test, Y_test)\n",
    "    print(\"\\n--------------------\\nTotal Testing Loss: \", score[1], \"\\nTesting Accuracy: \", score[1])\n",
    "    return training_history"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [],
   "source": [
    "def plot_training_history(training_history):\n",
    "    f, axarr = plt.subplots(2, sharex=True)\n",
    "    axarr[0].set_xlabel('Epoch')\n",
    "    axarr[0].set_ylabel('Training Accuracy')\n",
    "    axarr[0].plot(list(range(1, NP_EPOCH+1)), training_history.history['acc'])\n",
    "    axarr[1].set_ylabel('Training Error')\n",
    "    axarr[1].plot(list(range(1, NP_EPOCH+1)), training_history.history['loss'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "1s - loss: 0.0903 - acc: 0.1105\n",
      "Epoch 2/20\n",
      "1s - loss: 0.0899 - acc: 0.1290\n",
      "Epoch 3/20\n",
      "1s - loss: 0.0898 - acc: 0.1360\n",
      "Epoch 4/20\n",
      "1s - loss: 0.0897 - acc: 0.1633\n",
      "Epoch 5/20\n",
      "1s - loss: 0.0897 - acc: 0.1524\n",
      "Epoch 6/20\n",
      "1s - loss: 0.0896 - acc: 0.1815\n",
      "Epoch 7/20\n",
      "1s - loss: 0.0895 - acc: 0.2218\n",
      "Epoch 8/20\n",
      "1s - loss: 0.0894 - acc: 0.1716\n",
      "Epoch 9/20\n",
      "1s - loss: 0.0894 - acc: 0.2083\n",
      "Epoch 10/20\n",
      "2s - loss: 0.0893 - acc: 0.2450\n",
      "Epoch 11/20\n",
      "2s - loss: 0.0892 - acc: 0.2524\n",
      "Epoch 12/20\n",
      "2s - loss: 0.0891 - acc: 0.2671\n",
      "Epoch 13/20\n",
      "1s - loss: 0.0891 - acc: 0.2819\n",
      "Epoch 14/20\n",
      "2s - loss: 0.0890 - acc: 0.2603\n",
      "Epoch 15/20\n",
      "1s - loss: 0.0889 - acc: 0.2808\n",
      "Epoch 16/20\n",
      "1s - loss: 0.0888 - acc: 0.2997\n",
      "Epoch 17/20\n",
      "2s - loss: 0.0887 - acc: 0.3098\n",
      "Epoch 18/20\n",
      "1s - loss: 0.0887 - acc: 0.3099\n",
      "Epoch 19/20\n",
      "2s - loss: 0.0886 - acc: 0.3151\n",
      "Epoch 20/20\n",
      "2s - loss: 0.0885 - acc: 0.2956\n",
      "1000/1000 [==============================] - 0s     \b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\n",
      "\n",
      "--------------------\n",
      "Total Testing Loss:  0.306 \n",
      "Testing Accuracy:  0.306\n",
      "CPU times: user 1min 23s, sys: 3.32 s, total: 1min 27s\n",
      "Wall time: 39.5 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "history = train_and_show_result(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAD8CAYAAABdCyJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12f7b1358>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_training_history(history)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Cross Entropy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer=SGD(lr=0.1),\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "2s - loss: 2.2582 - acc: 0.1818\n",
      "Epoch 2/20\n",
      "2s - loss: 1.8396 - acc: 0.4324\n",
      "Epoch 3/20\n",
      "2s - loss: 1.3142 - acc: 0.6353\n",
      "Epoch 4/20\n",
      "2s - loss: 0.9522 - acc: 0.7361\n",
      "Epoch 5/20\n",
      "1s - loss: 0.7551 - acc: 0.7971\n",
      "Epoch 6/20\n",
      "1s - loss: 0.6333 - acc: 0.8351\n",
      "Epoch 7/20\n",
      "2s - loss: 0.5564 - acc: 0.8488\n",
      "Epoch 8/20\n",
      "2s - loss: 0.5057 - acc: 0.8604\n",
      "Epoch 9/20\n",
      "2s - loss: 0.4641 - acc: 0.8732\n",
      "Epoch 10/20\n",
      "1s - loss: 0.4362 - acc: 0.8802\n",
      "Epoch 11/20\n",
      "1s - loss: 0.4139 - acc: 0.8864\n",
      "Epoch 12/20\n",
      "1s - loss: 0.3939 - acc: 0.8905\n",
      "Epoch 13/20\n",
      "2s - loss: 0.3821 - acc: 0.8932\n",
      "Epoch 14/20\n",
      "1s - loss: 0.3699 - acc: 0.8964\n",
      "Epoch 15/20\n",
      "1s - loss: 0.3587 - acc: 0.8999\n",
      "Epoch 16/20\n",
      "1s - loss: 0.3498 - acc: 0.9006\n",
      "Epoch 17/20\n",
      "1s - loss: 0.3425 - acc: 0.9012\n",
      "Epoch 18/20\n",
      "1s - loss: 0.3342 - acc: 0.9042\n",
      "Epoch 19/20\n",
      "1s - loss: 0.3266 - acc: 0.9079\n",
      "Epoch 20/20\n",
      "1s - loss: 0.3236 - acc: 0.9080\n",
      " 960/1000 [===========================>..] - ETA: 0s\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\n",
      "--------------------\n",
      "Total Testing Loss:  0.88 \n",
      "Testing Accuracy:  0.88\n",
      "CPU times: user 1min 25s, sys: 3.68 s, total: 1min 29s\n",
      "Wall time: 39.5 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "history = train_and_show_result(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAD8CAYAAACYebj1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nmIT/DzW+IyzG+ol6AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12415db00>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_training_history(history)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Deeper"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "\n",
    "model.add(Dense(input_dim=28*28, output_dim=500, activation='sigmoid'))\n",
    "\n",
    "model.add(Dense(output_dim=500, activation='sigmoid'))\n",
    "model.add(Dense(output_dim=500, activation='sigmoid'))\n",
    "model.add(Dense(output_dim=500, activation='sigmoid'))\n",
    "model.add(Dense(output_dim=500, activation='sigmoid'))\n",
    "model.add(Dense(output_dim=500, activation='sigmoid'))\n",
    "model.add(Dense(output_dim=500, activation='sigmoid'))\n",
    "model.add(Dense(output_dim=500, activation='sigmoid'))\n",
    "model.add(Dense(output_dim=500, activation='sigmoid'))\n",
    "\n",
    "model.add(Dense(output_dim=10, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [],
   "source": [
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer=SGD(lr=0.1),\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "8s - loss: 2.4371 - acc: 0.0981\n",
      "Epoch 2/20\n",
      "6s - loss: 2.3626 - acc: 0.1003\n",
      "Epoch 3/20\n",
      "6s - loss: 2.3477 - acc: 0.0997\n",
      "Epoch 4/20\n",
      "6s - loss: 2.3375 - acc: 0.0985\n",
      "Epoch 5/20\n",
      "7s - loss: 2.3319 - acc: 0.1018\n",
      "Epoch 6/20\n",
      "8s - loss: 2.3272 - acc: 0.1037\n",
      "Epoch 7/20\n",
      "7s - loss: 2.3263 - acc: 0.1033\n",
      "Epoch 8/20\n",
      "6s - loss: 2.3268 - acc: 0.0984\n",
      "Epoch 9/20\n",
      "6s - loss: 2.3187 - acc: 0.1015\n",
      "Epoch 10/20\n",
      "7s - loss: 2.3181 - acc: 0.1089\n",
      "Epoch 11/20\n",
      "6s - loss: 2.3142 - acc: 0.1018\n",
      "Epoch 12/20\n",
      "7s - loss: 2.3146 - acc: 0.1029\n",
      "Epoch 13/20\n",
      "6s - loss: 2.3137 - acc: 0.1016\n",
      "Epoch 14/20\n",
      "6s - loss: 2.3120 - acc: 0.1058\n",
      "Epoch 15/20\n",
      "6s - loss: 2.3124 - acc: 0.1032\n",
      "Epoch 16/20\n",
      "6s - loss: 2.3107 - acc: 0.1045\n",
      "Epoch 17/20\n",
      "6s - loss: 2.3109 - acc: 0.1012\n",
      "Epoch 18/20\n",
      "7s - loss: 2.3093 - acc: 0.1042\n",
      "Epoch 19/20\n",
      "6s - loss: 2.3095 - acc: 0.1023\n",
      "Epoch 20/20\n",
      "6s - loss: 2.3090 - acc: 0.1029\n",
      " 992/1000 [============================>.] - ETA: 0s\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\n",
      "--------------------\n",
      "Total Testing Loss:  0.126 \n",
      "Testing Accuracy:  0.126\n",
      "CPU times: user 5min 38s, sys: 21.8 s, total: 5min 59s\n",
      "Wall time: 2min 22s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "history = train_and_show_result(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAD8CAYAAABdCyJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x130a1b128>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_training_history(history)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# ReLU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "\n",
    "model.add(Dense(input_dim=28*28, output_dim=500, activation='relu'))\n",
    "\n",
    "model.add(Dense(output_dim=500, activation='relu'))\n",
    "model.add(Dense(output_dim=500, activation='relu'))\n",
    "model.add(Dense(output_dim=500, activation='relu'))\n",
    "model.add(Dense(output_dim=500, activation='relu'))\n",
    "model.add(Dense(output_dim=500, activation='relu'))\n",
    "model.add(Dense(output_dim=500, activation='relu'))\n",
    "model.add(Dense(output_dim=500, activation='relu'))\n",
    "model.add(Dense(output_dim=500, activation='relu'))\n",
    "\n",
    "model.add(Dense(output_dim=10, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [],
   "source": [
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer=SGD(lr=0.1),\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "8s - loss: 1.9295 - acc: 0.3481\n",
      "Epoch 2/20\n",
      "8s - loss: 0.7868 - acc: 0.7438\n",
      "Epoch 3/20\n",
      "8s - loss: 0.4664 - acc: 0.8639\n",
      "Epoch 4/20\n",
      "6s - loss: 0.2514 - acc: 0.9277\n",
      "Epoch 5/20\n",
      "6s - loss: 0.1565 - acc: 0.9554\n",
      "Epoch 6/20\n",
      "7s - loss: 0.1149 - acc: 0.9674\n",
      "Epoch 7/20\n",
      "6s - loss: 0.0878 - acc: 0.9755\n",
      "Epoch 8/20\n",
      "6s - loss: 0.0663 - acc: 0.9801\n",
      "Epoch 9/20\n",
      "6s - loss: 0.0376 - acc: 0.9897\n",
      "Epoch 10/20\n",
      "5s - loss: 0.0293 - acc: 0.9919\n",
      "Epoch 11/20\n",
      "5s - loss: 0.0157 - acc: 0.9966\n",
      "Epoch 12/20\n",
      "5s - loss: 0.0986 - acc: 0.9781\n",
      "Epoch 13/20\n",
      "5s - loss: 0.0174 - acc: 0.9952\n",
      "Epoch 14/20\n",
      "6s - loss: 0.0062 - acc: 0.9990\n",
      "Epoch 15/20\n",
      "5s - loss: 0.0021 - acc: 0.9998\n",
      "Epoch 16/20\n",
      "5s - loss: 0.0013 - acc: 0.9997\n",
      "Epoch 17/20\n",
      "5s - loss: 7.3086e-04 - acc: 0.9999\n",
      "Epoch 18/20\n",
      "6s - loss: 4.8557e-04 - acc: 1.0000\n",
      "Epoch 19/20\n",
      "5s - loss: 3.6850e-04 - acc: 1.0000\n",
      "Epoch 20/20\n",
      "6s - loss: 3.0188e-04 - acc: 1.0000\n",
      "1000/1000 [==============================] - 0s     \b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\n",
      "\n",
      "--------------------\n",
      "Total Testing Loss:  0.952 \n",
      "Testing Accuracy:  0.952\n",
      "CPU times: user 5min 22s, sys: 20.7 s, total: 5min 43s\n",
      "Wall time: 2min 8s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "history = train_and_show_result(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAD8CAYAAACYebj1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nXwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11f4a1208>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_training_history(history)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Adam (Adaptive Learning Rate + Momentum)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "-"
    }
   },
   "outputs": [],
   "source": [
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer=Adam(),\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "8s - loss: 0.2683 - acc: 0.9429\n",
      "Epoch 2/20\n",
      "8s - loss: 0.0823 - acc: 0.9769\n",
      "Epoch 3/20\n",
      "7s - loss: 0.0709 - acc: 0.9799\n",
      "Epoch 4/20\n",
      "7s - loss: 0.0707 - acc: 0.9803\n",
      "Epoch 5/20\n",
      "7s - loss: 0.0867 - acc: 0.9776\n",
      "Epoch 6/20\n",
      "8s - loss: 0.0599 - acc: 0.9849\n",
      "Epoch 7/20\n",
      "7s - loss: 0.0529 - acc: 0.9858\n",
      "Epoch 8/20\n",
      "7s - loss: 0.0515 - acc: 0.9869\n",
      "Epoch 9/20\n",
      "7s - loss: 0.0535 - acc: 0.9860\n",
      "Epoch 10/20\n",
      "8s - loss: 0.0547 - acc: 0.9858\n",
      "Epoch 11/20\n",
      "7s - loss: 0.0602 - acc: 0.9872\n",
      "Epoch 12/20\n",
      "7s - loss: 0.0490 - acc: 0.9893\n",
      "Epoch 13/20\n",
      "7s - loss: 0.0372 - acc: 0.9905\n",
      "Epoch 14/20\n",
      "7s - loss: 0.0245 - acc: 0.9944\n",
      "Epoch 15/20\n",
      "7s - loss: 0.0340 - acc: 0.9917\n",
      "Epoch 16/20\n",
      "7s - loss: 0.0296 - acc: 0.9928\n",
      "Epoch 17/20\n",
      "7s - loss: 0.0330 - acc: 0.9922\n",
      "Epoch 18/20\n",
      "7s - loss: 0.0510 - acc: 0.9883\n",
      "Epoch 19/20\n",
      "7s - loss: 0.0319 - acc: 0.9926\n",
      "Epoch 20/20\n",
      "7s - loss: 0.0376 - acc: 0.9912\n",
      " 992/1000 [============================>.] - ETA: 0s\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\n",
      "--------------------\n",
      "Total Testing Loss:  0.926 \n",
      "Testing Accuracy:  0.926\n",
      "CPU times: user 6min 59s, sys: 50 s, total: 7min 49s\n",
      "Wall time: 2min 38s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "history = train_and_show_result(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYwAAAD8CAYAAABkbJM/AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11d6ff518>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_training_history(history)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": true,
    "slideshow": {
     "slide_type": "-"
    }
   },
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "\n",
    "model.add(Dense(input_dim=28*28, output_dim=500, activation='relu'))\n",
    "\n",
    "model.add(Dropout(0.8))\n",
    "model.add(Dense(output_dim=500, activation='relu'))\n",
    "model.add(Dropout(0.8))\n",
    "model.add(Dense(output_dim=500, activation='relu'))\n",
    "model.add(Dropout(0.8))\n",
    "model.add(Dense(output_dim=500, activation='relu'))\n",
    "model.add(Dropout(0.8))\n",
    "model.add(Dense(output_dim=500, activation='relu'))\n",
    "model.add(Dropout(0.8))\n",
    "model.add(Dense(output_dim=500, activation='relu'))\n",
    "model.add(Dropout(0.8))\n",
    "model.add(Dense(output_dim=500, activation='relu'))\n",
    "model.add(Dropout(0.8))\n",
    "model.add(Dense(output_dim=500, activation='relu'))\n",
    "model.add(Dropout(0.8))\n",
    "model.add(Dense(output_dim=500, activation='relu'))\n",
    "\n",
    "model.add(Dense(output_dim=10, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [],
   "source": [
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer=Adam(),\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "9s - loss: 4.8432 - acc: 0.1033\n",
      "Epoch 2/20\n",
      "8s - loss: 2.3042 - acc: 0.1277\n",
      "Epoch 3/20\n",
      "8s - loss: 2.1631 - acc: 0.2094\n",
      "Epoch 4/20\n",
      "8s - loss: 2.0366 - acc: 0.2359\n",
      "Epoch 5/20\n",
      "8s - loss: 1.9533 - acc: 0.2450\n",
      "Epoch 6/20\n",
      "10s - loss: 1.8826 - acc: 0.2507\n",
      "Epoch 7/20\n",
      "8s - loss: 1.8410 - acc: 0.2527\n",
      "Epoch 8/20\n",
      "8s - loss: 1.8115 - acc: 0.2612\n",
      "Epoch 9/20\n",
      "9s - loss: 1.7930 - acc: 0.2537\n",
      "Epoch 10/20\n",
      "12s - loss: 1.7721 - acc: 0.2645\n",
      "Epoch 11/20\n",
      "9s - loss: 1.7390 - acc: 0.2701\n",
      "Epoch 12/20\n",
      "8s - loss: 1.7319 - acc: 0.2687\n",
      "Epoch 13/20\n",
      "11s - loss: 1.7138 - acc: 0.2745\n",
      "Epoch 14/20\n",
      "10s - loss: 1.6979 - acc: 0.2792\n",
      "Epoch 15/20\n",
      "9s - loss: 1.7011 - acc: 0.2844\n",
      "Epoch 16/20\n",
      "8s - loss: 1.6855 - acc: 0.2812\n",
      "Epoch 17/20\n",
      "8s - loss: 1.6703 - acc: 0.2888\n",
      "Epoch 18/20\n",
      "9s - loss: 1.6623 - acc: 0.2925\n",
      "Epoch 19/20\n",
      "8s - loss: 1.6514 - acc: 0.2930\n",
      "Epoch 20/20\n",
      "7s - loss: 1.6401 - acc: 0.2999\n",
      "1000/1000 [==============================] - 0s     \b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\n",
      "\n",
      "--------------------\n",
      "Total Testing Loss:  0.099 \n",
      "Testing Accuracy:  0.099\n",
      "CPU times: user 7min 25s, sys: 1min 1s, total: 8min 27s\n",
      "Wall time: 3min 7s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "history = train_and_show_result(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAD8CAYAAACYebj1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\niP8P3WWNR0I6N0MAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x131e3a0f0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_training_history(history)"
   ]
  }
 ],
 "metadata": {
  "celltoolbar": "Slideshow",
  "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": "3.6.0"
  },
  "livereveal": {
   "scroll": true
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}