Keras.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "# Introducing Keras\n",
    "\n",
    "Be sure to \"pip install keras\" first!\n",
    "\n",
    "Keras is a layer on top of TensorFlow that makes things a lot easier. Not only is it easier to use, it's easier to tune.\n",
    "\n",
    "Let's set up the same deep neural network we set up with TensorFlow to learn from the MNIST data set.\n",
    "\n",
    "First we'll import all the stuff we need, which will initialize Keras as a side effect:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import keras\n",
    "from keras.datasets import mnist\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout\n",
    "from keras.optimizers import RMSprop"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "We'll load up the MNIST data set. In Keras, it's a little bit different - there are 60K training samples and 10K test samples. No \"validation\" samples."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "(mnist_train_images, mnist_train_labels), (mnist_test_images, mnist_test_labels) = mnist.load_data()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "We need to explicitly convert the data into the format Keras / TensorFlow expects. We divide the image data by 255 in order to normalize it into 0-1 range, after converting it into floating point values."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "train_images = mnist_train_images.reshape(60000, 784)\n",
    "test_images = mnist_test_images.reshape(10000, 784)\n",
    "train_images = train_images.astype('float32')\n",
    "test_images = test_images.astype('float32')\n",
    "train_images /= 255\n",
    "test_images /= 255"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "Now we'll convert the 0-9 labels into \"one-hot\" format, as we did for TensorFlow."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "train_labels = keras.utils.to_categorical(mnist_train_labels, 10)\n",
    "test_labels = keras.utils.to_categorical(mnist_test_labels, 10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "Let's take a peek at one of the training images just to make sure it looks OK:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 0.  0.  0.  1.  0.  0.  0.  0.  0.  0.]\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nWaL+HxAzIkDaMQ+AAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a55ad08d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "def display_sample(num):\n",
    "    #Print the one-hot array of this sample's label \n",
    "    print(train_labels[num])  \n",
    "    #Print the label converted back to a number\n",
    "    label = train_labels[num].argmax(axis=0)\n",
    "    #Reshape the 768 values to a 28x28 image\n",
    "    image = train_images[num].reshape([28,28])\n",
    "    plt.title('Sample: %d  Label: %d' % (num, label))\n",
    "    plt.imshow(image, cmap=plt.get_cmap('gray_r'))\n",
    "    plt.show()\n",
    "    \n",
    "display_sample(1234)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "Here's where things get exciting. All that code we wrote in Tensorflow creating placeholders, variables, and defining a bunch of linear algebra for each layer in our neural network? None of that is necessary with Keras!\n",
    "\n",
    "We can set up the same layers like this. The input layer of 784 features feeds into a ReLU layer of 512 nodes, which then goes into 10 nodes with softmax applied. Couldn't be simpler:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "model.add(Dense(512, activation='relu', input_shape=(784,)))\n",
    "model.add(Dense(10, activation='softmax'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "We can even get a nice description of the resulting model:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_1 (Dense)              (None, 512)               401920    \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                5130      \n",
      "=================================================================\n",
      "Total params: 407,050\n",
      "Trainable params: 407,050\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "Setting up our optimizer and loss function is just as simple. We will use the RMSProp optimizer here. Other choices include Adagrad, SGD, Adam, Adamax, and Nadam. See https://keras.io/optimizers/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer=RMSprop(),\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "Training our model is also just one line of code with Keras. Here we'll do 10 epochs with a batch size of 100. Keras is slower, and if we're not running on top of a GPU-accelerated Tensorflow this can take a fair amount of time (that's why I've limited it to just 10 epochs.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/10\n",
      "4s - loss: 0.2423 - acc: 0.9289 - val_loss: 0.1288 - val_acc: 0.9608\n",
      "Epoch 2/10\n",
      "3s - loss: 0.0985 - acc: 0.9707 - val_loss: 0.0898 - val_acc: 0.9725\n",
      "Epoch 3/10\n",
      "3s - loss: 0.0648 - acc: 0.9804 - val_loss: 0.0847 - val_acc: 0.9740\n",
      "Epoch 4/10\n",
      "3s - loss: 0.0479 - acc: 0.9856 - val_loss: 0.0663 - val_acc: 0.9809\n",
      "Epoch 5/10\n",
      "3s - loss: 0.0368 - acc: 0.9889 - val_loss: 0.0670 - val_acc: 0.9799\n",
      "Epoch 6/10\n",
      "3s - loss: 0.0271 - acc: 0.9921 - val_loss: 0.0713 - val_acc: 0.9798\n",
      "Epoch 7/10\n",
      "4s - loss: 0.0214 - acc: 0.9935 - val_loss: 0.0725 - val_acc: 0.9796\n",
      "Epoch 8/10\n",
      "3s - loss: 0.0166 - acc: 0.9954 - val_loss: 0.0773 - val_acc: 0.9802\n",
      "Epoch 9/10\n",
      "3s - loss: 0.0124 - acc: 0.9963 - val_loss: 0.0731 - val_acc: 0.9800\n",
      "Epoch 10/10\n",
      "3s - loss: 0.0098 - acc: 0.9972 - val_loss: 0.0712 - val_acc: 0.9813\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(train_images, train_labels,\n",
    "                    batch_size=100,\n",
    "                    epochs=10,\n",
    "                    verbose=2,\n",
    "                    validation_data=(test_images, test_labels))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "But, even with just 10 epochs, we've outperformed our Tensorflow version considerably!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test loss: 0.0711936488321\n",
      "Test accuracy: 0.9813\n"
     ]
    }
   ],
   "source": [
    "score = model.evaluate(test_images, test_labels, verbose=0)\n",
    "print('Test loss:', score[0])\n",
    "print('Test accuracy:', score[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "As before let's visualize the ones it got wrong. As this model is much better, we'll have to search deeper to find mistakes to look at."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a652cac18>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nB+sBcAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a6556c390>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAEgVJREFUeJzt3X/wXfOdx/Hnq6iGqBX5boQmUm3WMC1RX+wMpTatX2tL\nTNEwJEalq91uO6uskXQS2y3RaYpRWiEIStlNrDCs9atj2Ay+CCJZgklK5Mc3TW0TjArv/eOeb/fm\n63vPvbm/zv3m83rM3Pmee97nxzsn39c995xz7/coIjCz9Hyi6AbMrBgOv1miHH6zRDn8Zoly+M0S\n5fCbJcrhL5ikMZJC0rbZ8/slTapjOaMlbZS0TfO7bD9JkyU93u55U+Lw10DScknvZeFaI+kmSUNb\nsa6IODYi5tbY01fL5vtdRAyNiA9b0VfZevteZMofIem8CtPPkHRrK3tqhKS/knS3pF5J6yU9IGnv\novtqB4e/dn8XEUOBLwHdwLT+E6hkq96mZS8yQ7Pt8UXgI2Bewa3V6y+ABcDewAjgKeDuQjtqk636\nF7UVImIlcD/wBQBJv5X0E0lPAO8Ce0naWdIcSaskrZT0r31vxyVtI+lnktZJeh342/LlZ8v7Vtnz\ncyQtlbRB0hJJX5J0CzAauCfb814wwOHD7pIWZHuzVyWdU7bMGZLulHRzttyXJHXXuUnOBB6LiOVb\nOqOkCyW9VvZvm/DxSfQLSf8r6X8kjS8rVNzGWyIinoqIORGxPiI+AC4H9pa065Yua7Bx+LeQpFHA\nccBzZaPPAKYAOwErgJuATcDngQOAo4C+QJ8DHJ+N7wa+kbOuk4EZlAL2aeDrwO8j4gzgd2TvRiLi\npwPM/hvgTWD3bB2XSPqbsvrXs2n69ny/KFvvNZKuyd8SpWRmvVU9TKngNeDLwM7AxcCtkkaW1Q/J\nphkOTAfmSxqW1W6i8jbu3+e9ki6ssafDgdUR8fst+6cMQhHhR5UHsBzYCLxNKdzXAEOy2m+Bfymb\ndgTwfl89GzcReDQbfgT4+7LaUUAA25Yt71vZ8APA93N6+mrZ8zF9ywFGAR8CO5XVLwVuyoZnAA+V\n1fYF3qtju3w52y5Dc6aZAdxa4/IWASdkw5OBtwCV1Z+i9EJbbRtPBh6v49/zGWAlMLHo37l2PLat\n/WUieSdGxEMVam+UDe8JbAesKu0YgdI7rL5pdu83/YqcdY6itOfbUrsD6yNiQ7/1lL+1X102/C7w\nKUnbRsSmLVjPJGBeRGyso0cknQn8E6UXLoChlPbyfVZGlsrMCkr/tmrbuJ5euoD/Aq6JiNvrXc5g\n4vA3R/kv6BuU9krDKwRpFaVQ9xmds9w3gM/VsM7+3gKGSdqp7AVgNKW9WlNIGgKcDPQ/Tq91/j2B\n64DxwMKI+FDSIkBlk+0hSWUvAKMpHaJU28Zb2ssulIK/ICJ+0ujyBgsf8zdZRKyi9Is0S9KnJX1C\n0uckHZFNcifwj5I+k/3S5R2LXg/8UNKB2ZWEz2ehAVgD7FWhhzeA/wYulfQpSfsBZwPNvOQ2AfgD\n8GgN034i66PvsT2wI6UXsF4ASWeRnUQt85eUttV22fmPfYD7atjGNZP0aUqHV09ERK3nBbYKDn9r\nnAl8ElhCKSD/DvSdyLqO0i/b88CzwPxKC4mIfwN+AtwGbAD+A+g74XUpME3S25J+OMDsEym9nX4L\nuAuYnnPYshlJv5L0qyqTTQJu6fe2vJKJwHtlj9ciYgkwC1hI6YXsi8AT/eZ7EhgLrKO0Hb4R/38i\nLm8b9//33C/pogq9TQAOAs7S5p9dyHtHtlVQbf93Zra18Z7fLFEOv1miHH6zRDn8Zolq63X+4cOH\nx5gxY9q5SrOkLF++nHXr1qn6lA2GX9IxwJXANsD1ETEzb/oxY8bQ09PTyCrNLEd3d+3fz6r7bX/2\nDaqrgWMpfTZ8oqR9612embVXI8f8BwOvRsTrEfEnSt8QO6E5bZlZqzUS/j3Y/IsUb2bjNiNpiqQe\nST29vb0NrM7MmqnlZ/sjYnZEdEdEd1dXV6tXZ2Y1aiT8K9n822l934U2s0GgkfA/DYyV9FlJnwS+\nSenrlmY2CNR9qS8iNkn6B0rfUNsGuCEiXmpaZ2bWUg1d54+I+4D7mtSLmbWRP95rliiH3yxRDr9Z\nohx+s0Q5/GaJcvjNEuXwmyXK4TdLlMNvliiH3yxRDr9Zohx+s0Q5/GaJcvjNEuXwmyXK4TdLlMNv\nliiH3yxRDr9Zohx+s0Q5/GaJcvjNEuXwmyXK4TdLlMNvliiH3yxRDr9Zohx+s0Q5/GaJcvjNEtXQ\nLbolLQc2AB8CmyKiuxlNmVnrNRT+zJERsa4JyzGzNvLbfrNENRr+AB6S9IykKQNNIGmKpB5JPb29\nvQ2uzsyapdHwHxYR44Bjge9KOrz/BBExOyK6I6K7q6urwdWZWbM0FP6IWJn9XAvcBRzcjKbMrPXq\nDr+kHSXt1DcMHAUsblZjZtZajZztHwHcJalvObdFxH82pSvbzNtvv51bX7ZsWcXabbfd1tC6r7ji\nitx69v/fErvttltufeHChbn1Pffcs5ntbHXqDn9EvA7s38RezKyNfKnPLFEOv1miHH6zRDn8Zoly\n+M0S1Ywv9liDbr311tz6JZdcklt/+eWXm9nOZqpdytt///wLPh988EHF2tKlS3PnXbNmTW599erV\nuXVf6svnPb9Zohx+s0Q5/GaJcvjNEuXwmyXK4TdLlMNvlihf52+Dal+rPffcc3Pr7777bm592LBh\nFWsnnXRS7rzVrtMffvjH/jjTZqpdS9+0aVPF2qhRo3Lnfe+993Lr1bbrIYcckltPnff8Zoly+M0S\n5fCbJcrhN0uUw2+WKIffLFEOv1mifJ2/Capdh7/++utz6wceeGBufdq0abn1Qw89tGJtyJAhufO2\nWt61+kb/7PfJJ5/c0Pyp857fLFEOv1miHH6zRDn8Zoly+M0S5fCbJcrhN0uUr/M3wQ477JBbf+SR\nR9rUSeeZNWtWxVq1z0eMHTs2t77PPvvU1ZOVVN3zS7pB0lpJi8vGDZP0oKRl2c9dWtummTVbLW/7\nbwKO6TfuQuDhiBgLPJw9N7NBpGr4I+IxYH2/0ScAc7PhucCJTe7LzFqs3hN+IyJiVTa8GhhRaUJJ\nUyT1SOrp7e2tc3Vm1mwNn+2PiAAipz47Irojorurq6vR1ZlZk9Qb/jWSRgJkP9c2ryUza4d6w78A\nmJQNTwLubk47ZtYuVa/zS7od+AowXNKbwHRgJnCnpLOBFcAprWzSOtfTTz+dW7/sssvqXna1+xns\nuuuudS/bagh/REysUBrf5F7MrI388V6zRDn8Zoly+M0S5fCbJcrhN0uUv9JruT766KPc+gMPPJBb\nz/va7s4775w775FHHplbt8Z4z2+WKIffLFEOv1miHH6zRDn8Zoly+M0S5fCbJcrX+S3XnDlzcuvT\np0+ve9kzZ87Mre+33351L9uq857fLFEOv1miHH6zRDn8Zoly+M0S5fCbJcrhN0uUr/Nbrnvvvbeh\n+UePHl2xNmnSpIo1az3v+c0S5fCbJcrhN0uUw2+WKIffLFEOv1miHH6zRPk6f+Kee+653Po999yT\nW5eUWz///PMr1rbffvvcea21qu75Jd0gaa2kxWXjZkhaKWlR9jiutW2aWbPV8rb/JuCYAcZfHhHj\nssd9zW3LzFqtavgj4jFgfRt6MbM2auSE3/ckvZAdFuxSaSJJUyT1SOrp7e1tYHVm1kz1hv+XwF7A\nOGAVMKvShBExOyK6I6K7q6urztWZWbPVFf6IWBMRH0bER8B1wMHNbcvMWq2u8EsaWfZ0ArC40rRm\n1pmqXueXdDvwFWC4pDeB6cBXJI0DAlgOfLuFPVoD3nnnndz6jBkzcusRkVsfP358bv073/lObt2K\nUzX8ETFxgNH5d3Iws47nj/eaJcrhN0uUw2+WKIffLFEOv1mi/JXerdyNN96YW6/2p7mHDBmSWz/r\nrLO2uCfrDN7zmyXK4TdLlMNvliiH3yxRDr9Zohx+s0Q5/GaJ8nX+rcCyZcsq1qZOndrQsvP+9DbA\naaed1tDyrTje85slyuE3S5TDb5Yoh98sUQ6/WaIcfrNEOfxmifJ1/kGg2p/PvvTSSyvWNm7c2NC6\njz/++Ibmt87lPb9Zohx+s0Q5/GaJcvjNEuXwmyXK4TdLlMNvlqhabtE9CrgZGEHpltyzI+JKScOA\nO4AxlG7TfUpE/KF1raZr3rx5ufW5c+fWvezJkyfn1g866KC6l22drZY9/ybgvIjYF/hr4LuS9gUu\nBB6OiLHAw9lzMxskqoY/IlZFxLPZ8AZgKbAHcALQt8uZC5zYqibNrPm26Jhf0hjgAOBJYERErMpK\nqykdFpjZIFFz+CUNBeYBP4iIP5bXovTh8wE/gC5piqQeST29vb0NNWtmzVNT+CVtRyn4v46I+dno\nNZJGZvWRwNqB5o2I2RHRHRHdXV1dzejZzJqgavglCZgDLI2In5eVFgCTsuFJwN3Nb8/MWqWWr/Qe\nCpwBvChpUTbuImAmcKeks4EVwCmtadFeeeWVli172rRpLVt2NXfccUdu/dRTT21TJ2mqGv6IeBxQ\nhfL45rZjZu3iT/iZJcrhN0uUw2+WKIffLFEOv1miHH6zRPlPdw8CPT09dc/7ox/9KLc+evTo3Pr7\n77+fW58/f35u/cc//nHF2lVXXZU7r7WW9/xmiXL4zRLl8JslyuE3S5TDb5Yoh98sUQ6/WaJ8nX8Q\nWLhwYd3zrl+/Pre+ZMmS3Prpp5+eW1+xYkVuferUqRVrRxxxRO681lre85slyuE3S5TDb5Yoh98s\nUQ6/WaIcfrNEOfxmifJ1/kFgwoQJufVrr722Yu3qq6/OnbdavXQntsqmTJmSW7/gggty61Yc7/nN\nEuXwmyXK4TdLlMNvliiH3yxRDr9Zohx+s0RVvc4vaRRwMzACCGB2RFwpaQZwDtCbTXpRRNzXqkZT\ndvHFF+fWn3jiiYq1xYsX5847bty43Pq0adNy60cffXRu3TpXLR/y2QScFxHPStoJeEbSg1nt8oj4\nWevaM7NWqRr+iFgFrMqGN0haCuzR6sbMrLW26Jhf0hjgAODJbNT3JL0g6QZJu1SYZ4qkHkk9vb29\nA01iZgWoOfyShgLzgB9ExB+BXwJ7AeMovTOYNdB8ETE7Irojorurq6sJLZtZM9QUfknbUQr+ryNi\nPkBErImIDyPiI+A64ODWtWlmzVY1/JIEzAGWRsTPy8aPLJtsApB/WtnMOkotZ/sPBc4AXpS0KBt3\nETBR0jhKl/+WA99uSYdGtcOl559/vk2d2NaklrP9jwMaoORr+maDmD/hZ5Yoh98sUQ6/WaIcfrNE\nOfxmiXL4zRLl8JslyuE3S5TDb5Yoh98sUQ6/WaIcfrNEOfxmiXL4zRKlardgburKpF5gRdmo4cC6\ntjWwZTq1t07tC9xbvZrZ254RUdPfy2tr+D+2cqknIroLayBHp/bWqX2Be6tXUb35bb9Zohx+s0QV\nHf7ZBa8/T6f21ql9gXurVyG9FXrMb2bFKXrPb2YFcfjNElVI+CUdI+llSa9KurCIHiqRtFzSi5IW\nSeopuJcbJK2VtLhs3DBJD0palv0c8B6JBfU2Q9LKbNstknRcQb2NkvSopCWSXpL0/Wx8odsup69C\ntlvbj/klbQO8AnwNeBN4GpgYEUva2kgFkpYD3RFR+AdCJB0ObARujogvZON+CqyPiJnZC+cuEfHP\nHdLbDGBj0bdtz+4mNbL8tvLAicBkCtx2OX2dQgHbrYg9/8HAqxHxekT8CfgNcEIBfXS8iHgMWN9v\n9AnA3Gx4LqVfnrar0FtHiIhVEfFsNrwB6LutfKHbLqevQhQR/j2AN8qev0mBG2AAATwk6RlJU4pu\nZgAjImJVNrwaGFFkMwOoetv2dup3W/mO2Xb13O6+2XzC7+MOi4hxwLHAd7O3tx0pSsdsnXSttqbb\ntrfLALeV/7Mit129t7tvtiLCvxIYVfb8M9m4jhARK7Ofa4G76Lxbj6/pu0Ny9nNtwf38WSfdtn2g\n28rTAduuk253X0T4nwbGSvqspE8C3wQWFNDHx0jaMTsRg6QdgaPovFuPLwAmZcOTgLsL7GUznXLb\n9kq3lafgbddxt7uPiLY/gOMonfF/DZhaRA8V+toLeD57vFR0b8DtlN4GfkDp3MjZwK7Aw8Ay4CFg\nWAf1dgvwIvACpaCNLKi3wyi9pX8BWJQ9jit62+X0Vch288d7zRLlE35miXL4zRLl8JslyuE3S5TD\nb5Yoh98sUQ6/WaL+D+3w0/eklMQUAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a655a0e48>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a65999f98>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nJcrhN0uUw2+WKIffLFEOv1miHH6zRP0//pfgeEibeYcAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a655d2240>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nxQ9WDQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a659997f0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nh98sUQ6/WaIcfrNEOfxmifp/Sum7mblDrREAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a65587c50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a65a647f0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n4TdL1P8DgBNtc+CRy3IAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a65a23080>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nHH6zRP0fJ3jZk78BDSgAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a65cca7f0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nyuE3S9T/A/U7DbAv/a0mAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a65d3da20>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nb5ap/wPKmPj24K0ueAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a65cdee80>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nmXL4zTLl8JtlyuE3y5TDb5ap/wPYiM/x9CZRRQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a65d71fd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nmyXK4TdLlMNvlqj/A/s+sv5YqWQgAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a65da4b38>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a65608ef0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nuPF0I0dMAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a65ad6518>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAEICAYAAACQ6CLfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nUQ6/WaIcfrNEOfxmiXL4zRLl8Jsl6v8B/pLm84/0T8QAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x14a65d4d710>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for x in range(1000):\n",
    "    test_image = test_images[x,:].reshape(1,784)\n",
    "    predicted_cat = model.predict(test_image).argmax()\n",
    "    label = test_labels[x].argmax()\n",
    "    if (predicted_cat != label):\n",
    "        plt.title('Prediction: %d Label: %d' % (predicted_cat, label))\n",
    "        plt.imshow(test_image.reshape([28,28]), cmap=plt.get_cmap('gray_r'))\n",
    "        plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "source": [
    "You can see most of the ones it's having trouble with, are images a human would have trouble with as well!\n",
    "\n",
    "## Excercise\n",
    "\n",
    "As before, see if you can improve on the results! Does running more epochs help considerably? How about trying different optimizers?\n",
    "\n",
    "You can also take advantage of Keras's ease of use to try different topologies quickly. Keras includes a MNIST example, where they add an additional layer, and use Dropout at each step to prevent overfitting, like this:\n",
    "\n",
    "`\n",
    "model = Sequential()\n",
    "model.add(Dense(512, activation='relu', input_shape=(784,)))\n",
    "model.add(Dropout(0.2))\n",
    "model.add(Dense(512, activation='relu'))\n",
    "model.add(Dropout(0.2))\n",
    "model.add(Dense(10, activation='softmax'))\n",
    "`\n",
    "\n",
    "Try adapting that to our code above and see if it makes a difference or not."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}