mnist_cnn.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting MNIST_data/train-images-idx3-ubyte.gz\n",
      "Extracting MNIST_data/train-labels-idx1-ubyte.gz\n",
      "Extracting MNIST_data/t10k-images-idx3-ubyte.gz\n",
      "Extracting MNIST_data/t10k-labels-idx1-ubyte.gz\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "\n",
    "# Import MNIST data\n",
    "from tensorflow.examples.tutorials.mnist import input_data\n",
    "mnist = input_data.read_data_sets(\"MNIST_data/\", one_hot=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Parameters\n",
    "learning_rate = 0.001\n",
    "training_iters = 200000\n",
    "batch_size = 128\n",
    "display_step = 10\n",
    "\n",
    "# Network Parameters\n",
    "n_input = 784 # MNIST data input (img shape: 28*28)\n",
    "n_classes = 10 # MNIST total classes (0-9 digits)\n",
    "dropout = 0.75 # Dropout, probability to keep units\n",
    "\n",
    "# tf Graph input\n",
    "x = tf.placeholder(tf.float32, [None, n_input])\n",
    "y = tf.placeholder(tf.float32, [None, n_classes])\n",
    "keep_prob = tf.placeholder(tf.float32) #dropout (keep probability)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Create some wrappers for simplicity\n",
    "def conv2d(x, W, b, strides=1):\n",
    "    # Conv2D wrapper, with bias and relu activation\n",
    "    x = tf.nn.conv2d(x, W, strides=[1, strides, strides, 1], padding='SAME')\n",
    "    x = tf.nn.bias_add(x, b)\n",
    "    return tf.nn.relu(x)\n",
    "\n",
    "\n",
    "def maxpool2d(x, k=2):\n",
    "    # MaxPool2D wrapper\n",
    "    return tf.nn.max_pool(x, ksize=[1, k, k, 1], strides=[1, k, k, 1],\n",
    "                          padding='SAME')\n",
    "\n",
    "\n",
    "# Create model\n",
    "def conv_net(x, weights, biases, dropout):\n",
    "    # Reshape input picture\n",
    "    x = tf.reshape(x, shape=[-1, 28, 28, 1])\n",
    "\n",
    "    # Convolution Layer\n",
    "    conv1 = conv2d(x, weights['wc1'], biases['bc1'])\n",
    "    # Max Pooling (down-sampling)\n",
    "    conv1 = maxpool2d(conv1, k=2)\n",
    "\n",
    "    # Convolution Layer\n",
    "    conv2 = conv2d(conv1, weights['wc2'], biases['bc2'])\n",
    "    # Max Pooling (down-sampling)\n",
    "    conv2 = maxpool2d(conv2, k=2)\n",
    "\n",
    "    # Fully connected layer\n",
    "    # Reshape conv2 output to fit fully connected layer input\n",
    "    fc1 = tf.reshape(conv2, [-1, weights['wd1'].get_shape().as_list()[0]])\n",
    "    fc1 = tf.add(tf.matmul(fc1, weights['wd1']), biases['bd1'])\n",
    "    fc1 = tf.nn.relu(fc1)\n",
    "    # Apply Dropout\n",
    "    fc1 = tf.nn.dropout(fc1, dropout)\n",
    "\n",
    "    # Output, class prediction\n",
    "    out = tf.add(tf.matmul(fc1, weights['out']), biases['out'])\n",
    "    return out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Store layers weight & bias\n",
    "weights = {\n",
    "    # 5x5 conv, 1 input, 32 outputs\n",
    "    'wc1': tf.Variable(tf.random_normal([5, 5, 1, 32])),\n",
    "    # 5x5 conv, 32 inputs, 64 outputs\n",
    "    'wc2': tf.Variable(tf.random_normal([5, 5, 32, 64])),\n",
    "    # fully connected, 7*7*64 inputs, 1024 outputs\n",
    "    'wd1': tf.Variable(tf.random_normal([7*7*64, 1024])),\n",
    "    # 1024 inputs, 10 outputs (class prediction)\n",
    "    'out': tf.Variable(tf.random_normal([1024, n_classes]))\n",
    "}\n",
    "\n",
    "biases = {\n",
    "    'bc1': tf.Variable(tf.random_normal([32])),\n",
    "    'bc2': tf.Variable(tf.random_normal([64])),\n",
    "    'bd1': tf.Variable(tf.random_normal([1024])),\n",
    "    'out': tf.Variable(tf.random_normal([n_classes]))\n",
    "}\n",
    "\n",
    "# Construct model\n",
    "pred = conv_net(x, weights, biases, keep_prob)\n",
    "\n",
    "# Define loss and optimizer\n",
    "cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=pred, labels=y))\n",
    "optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost)\n",
    "\n",
    "# Evaluate model\n",
    "correct_pred = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))\n",
    "accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))\n",
    "\n",
    "# Initializing the variables\n",
    "init = tf.global_variables_initializer()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iter 1280, Minibatch Loss= 26379.152344, Training Accuracy= 0.17969\n",
      "Iter 2560, Minibatch Loss= 12574.867188, Training Accuracy= 0.47656\n",
      "Iter 3840, Minibatch Loss= 7169.190918, Training Accuracy= 0.67188\n",
      "Iter 5120, Minibatch Loss= 3932.330566, Training Accuracy= 0.82031\n",
      "Iter 6400, Minibatch Loss= 3236.889893, Training Accuracy= 0.85938\n",
      "Iter 7680, Minibatch Loss= 7344.635742, Training Accuracy= 0.75781\n",
      "Iter 8960, Minibatch Loss= 3225.331299, Training Accuracy= 0.83594\n",
      "Iter 10240, Minibatch Loss= 3658.146484, Training Accuracy= 0.82812\n",
      "Iter 11520, Minibatch Loss= 1743.686523, Training Accuracy= 0.92969\n",
      "Iter 12800, Minibatch Loss= 2878.284668, Training Accuracy= 0.82031\n",
      "Iter 14080, Minibatch Loss= 989.606445, Training Accuracy= 0.90625\n",
      "Iter 15360, Minibatch Loss= 1343.491211, Training Accuracy= 0.88281\n",
      "Iter 16640, Minibatch Loss= 1616.763062, Training Accuracy= 0.89844\n",
      "Iter 17920, Minibatch Loss= 568.510925, Training Accuracy= 0.92969\n",
      "Iter 19200, Minibatch Loss= 1207.645020, Training Accuracy= 0.89062\n",
      "Iter 20480, Minibatch Loss= 647.591064, Training Accuracy= 0.93750\n",
      "Iter 21760, Minibatch Loss= 2941.610840, Training Accuracy= 0.82031\n",
      "Iter 23040, Minibatch Loss= 959.511536, Training Accuracy= 0.96094\n",
      "Iter 24320, Minibatch Loss= 1336.018311, Training Accuracy= 0.92188\n",
      "Iter 25600, Minibatch Loss= 1138.427856, Training Accuracy= 0.92188\n",
      "Iter 26880, Minibatch Loss= 1571.023193, Training Accuracy= 0.90625\n",
      "Iter 28160, Minibatch Loss= 1107.148071, Training Accuracy= 0.90625\n",
      "Iter 29440, Minibatch Loss= 1035.199829, Training Accuracy= 0.92188\n",
      "Iter 30720, Minibatch Loss= 700.992676, Training Accuracy= 0.95312\n",
      "Iter 32000, Minibatch Loss= 1156.716309, Training Accuracy= 0.91406\n",
      "Iter 33280, Minibatch Loss= 442.507935, Training Accuracy= 0.91406\n",
      "Iter 34560, Minibatch Loss= 796.133179, Training Accuracy= 0.92188\n",
      "Iter 35840, Minibatch Loss= 565.179932, Training Accuracy= 0.98438\n",
      "Iter 37120, Minibatch Loss= 1150.137817, Training Accuracy= 0.92969\n",
      "Iter 38400, Minibatch Loss= 251.375259, Training Accuracy= 0.96094\n",
      "Iter 39680, Minibatch Loss= 227.087280, Training Accuracy= 0.96094\n",
      "Iter 40960, Minibatch Loss= 1390.556274, Training Accuracy= 0.92188\n",
      "Iter 42240, Minibatch Loss= 1042.870850, Training Accuracy= 0.89062\n",
      "Iter 43520, Minibatch Loss= 117.093872, Training Accuracy= 0.97656\n",
      "Iter 44800, Minibatch Loss= 250.684296, Training Accuracy= 0.99219\n",
      "Iter 46080, Minibatch Loss= 51.964508, Training Accuracy= 0.99219\n",
      "Iter 47360, Minibatch Loss= 689.591431, Training Accuracy= 0.94531\n",
      "Iter 48640, Minibatch Loss= 468.123596, Training Accuracy= 0.94531\n",
      "Iter 49920, Minibatch Loss= 628.555542, Training Accuracy= 0.92969\n",
      "Iter 51200, Minibatch Loss= 247.675461, Training Accuracy= 0.95312\n",
      "Iter 52480, Minibatch Loss= 137.807098, Training Accuracy= 0.96875\n",
      "Iter 53760, Minibatch Loss= 26.798828, Training Accuracy= 0.98438\n",
      "Iter 55040, Minibatch Loss= 1003.957520, Training Accuracy= 0.93750\n",
      "Iter 56320, Minibatch Loss= 815.027893, Training Accuracy= 0.92188\n",
      "Iter 57600, Minibatch Loss= 247.680893, Training Accuracy= 0.96875\n",
      "Iter 58880, Minibatch Loss= 707.751160, Training Accuracy= 0.94531\n",
      "Iter 60160, Minibatch Loss= 224.394165, Training Accuracy= 0.96875\n",
      "Iter 61440, Minibatch Loss= 650.173218, Training Accuracy= 0.93750\n",
      "Iter 62720, Minibatch Loss= 428.961853, Training Accuracy= 0.95312\n",
      "Iter 64000, Minibatch Loss= 246.220703, Training Accuracy= 0.97656\n",
      "Iter 65280, Minibatch Loss= 486.930542, Training Accuracy= 0.96094\n",
      "Iter 66560, Minibatch Loss= 972.463196, Training Accuracy= 0.93750\n",
      "Iter 67840, Minibatch Loss= 923.462952, Training Accuracy= 0.92188\n",
      "Iter 69120, Minibatch Loss= 457.299988, Training Accuracy= 0.96875\n",
      "Iter 70400, Minibatch Loss= 1077.916992, Training Accuracy= 0.96094\n",
      "Iter 71680, Minibatch Loss= 176.535049, Training Accuracy= 0.96094\n",
      "Iter 72960, Minibatch Loss= 446.493683, Training Accuracy= 0.96094\n",
      "Iter 74240, Minibatch Loss= 213.332153, Training Accuracy= 0.96094\n",
      "Iter 75520, Minibatch Loss= 314.190735, Training Accuracy= 0.96875\n",
      "Iter 76800, Minibatch Loss= 808.803101, Training Accuracy= 0.94531\n",
      "Iter 78080, Minibatch Loss= 830.059570, Training Accuracy= 0.94531\n",
      "Iter 79360, Minibatch Loss= 404.078918, Training Accuracy= 0.96875\n",
      "Iter 80640, Minibatch Loss= 92.632729, Training Accuracy= 0.97656\n",
      "Iter 81920, Minibatch Loss= 1104.021362, Training Accuracy= 0.95312\n",
      "Iter 83200, Minibatch Loss= 671.347168, Training Accuracy= 0.95312\n",
      "Iter 84480, Minibatch Loss= 430.457672, Training Accuracy= 0.94531\n",
      "Iter 85760, Minibatch Loss= 435.548889, Training Accuracy= 0.94531\n",
      "Iter 87040, Minibatch Loss= 356.230011, Training Accuracy= 0.95312\n",
      "Iter 88320, Minibatch Loss= 448.879059, Training Accuracy= 0.96094\n",
      "Iter 89600, Minibatch Loss= 150.658142, Training Accuracy= 0.99219\n",
      "Iter 90880, Minibatch Loss= 420.456085, Training Accuracy= 0.96094\n",
      "Iter 92160, Minibatch Loss= 1253.912842, Training Accuracy= 0.92188\n",
      "Iter 93440, Minibatch Loss= 421.552185, Training Accuracy= 0.96875\n",
      "Iter 94720, Minibatch Loss= 267.954102, Training Accuracy= 0.96875\n",
      "Iter 96000, Minibatch Loss= 335.716827, Training Accuracy= 0.96875\n",
      "Iter 97280, Minibatch Loss= 478.622070, Training Accuracy= 0.95312\n",
      "Iter 98560, Minibatch Loss= 221.933121, Training Accuracy= 0.96875\n",
      "Iter 99840, Minibatch Loss= 360.697388, Training Accuracy= 0.96094\n",
      "Iter 101120, Minibatch Loss= 565.152039, Training Accuracy= 0.96875\n",
      "Iter 102400, Minibatch Loss= 459.141937, Training Accuracy= 0.95312\n",
      "Iter 103680, Minibatch Loss= 253.776825, Training Accuracy= 0.97656\n",
      "Iter 104960, Minibatch Loss= 313.809875, Training Accuracy= 0.96875\n",
      "Iter 106240, Minibatch Loss= 332.932434, Training Accuracy= 0.94531\n",
      "Iter 107520, Minibatch Loss= 77.198624, Training Accuracy= 0.97656\n",
      "Iter 108800, Minibatch Loss= 418.387939, Training Accuracy= 0.93750\n",
      "Iter 110080, Minibatch Loss= 295.965363, Training Accuracy= 0.97656\n",
      "Iter 111360, Minibatch Loss= 78.409195, Training Accuracy= 0.99219\n",
      "Iter 112640, Minibatch Loss= 319.248108, Training Accuracy= 0.96094\n",
      "Iter 113920, Minibatch Loss= 121.613380, Training Accuracy= 0.99219\n",
      "Iter 115200, Minibatch Loss= 378.153625, Training Accuracy= 0.96875\n",
      "Iter 116480, Minibatch Loss= 223.808975, Training Accuracy= 0.95312\n",
      "Iter 117760, Minibatch Loss= 263.196045, Training Accuracy= 0.96094\n",
      "Iter 119040, Minibatch Loss= 243.524017, Training Accuracy= 0.96094\n",
      "Iter 120320, Minibatch Loss= 376.054443, Training Accuracy= 0.97656\n",
      "Iter 121600, Minibatch Loss= 483.301453, Training Accuracy= 0.94531\n",
      "Iter 122880, Minibatch Loss= 289.763367, Training Accuracy= 0.96094\n",
      "Iter 124160, Minibatch Loss= 68.436089, Training Accuracy= 0.98438\n",
      "Iter 125440, Minibatch Loss= 184.547348, Training Accuracy= 0.96875\n",
      "Iter 126720, Minibatch Loss= 477.810364, Training Accuracy= 0.92969\n",
      "Iter 128000, Minibatch Loss= 267.976257, Training Accuracy= 0.96875\n",
      "Iter 129280, Minibatch Loss= 444.874756, Training Accuracy= 0.96875\n",
      "Iter 130560, Minibatch Loss= 312.597412, Training Accuracy= 0.98438\n",
      "Iter 131840, Minibatch Loss= 0.000000, Training Accuracy= 1.00000\n",
      "Iter 133120, Minibatch Loss= 97.727928, Training Accuracy= 0.97656\n",
      "Iter 134400, Minibatch Loss= 169.662689, Training Accuracy= 0.95312\n",
      "Iter 135680, Minibatch Loss= 256.318237, Training Accuracy= 0.96875\n",
      "Iter 136960, Minibatch Loss= 218.205994, Training Accuracy= 0.95312\n",
      "Iter 138240, Minibatch Loss= 128.905411, Training Accuracy= 0.96875\n",
      "Iter 139520, Minibatch Loss= 171.237686, Training Accuracy= 0.98438\n",
      "Iter 140800, Minibatch Loss= 230.440948, Training Accuracy= 0.96875\n",
      "Iter 142080, Minibatch Loss= 245.098175, Training Accuracy= 0.95312\n",
      "Iter 143360, Minibatch Loss= 370.847839, Training Accuracy= 0.95312\n",
      "Iter 144640, Minibatch Loss= 80.452843, Training Accuracy= 0.96875\n",
      "Iter 145920, Minibatch Loss= 584.898193, Training Accuracy= 0.96094\n",
      "Iter 147200, Minibatch Loss= 380.209473, Training Accuracy= 0.94531\n",
      "Iter 148480, Minibatch Loss= 289.300476, Training Accuracy= 0.96094\n",
      "Iter 149760, Minibatch Loss= 216.289856, Training Accuracy= 0.94531\n",
      "Iter 151040, Minibatch Loss= 82.720680, Training Accuracy= 0.99219\n",
      "Iter 152320, Minibatch Loss= 351.038116, Training Accuracy= 0.96094\n",
      "Iter 153600, Minibatch Loss= 33.509094, Training Accuracy= 0.98438\n",
      "Iter 154880, Minibatch Loss= 18.992676, Training Accuracy= 0.99219\n",
      "Iter 156160, Minibatch Loss= 105.569962, Training Accuracy= 0.96094\n",
      "Iter 157440, Minibatch Loss= 256.662048, Training Accuracy= 0.97656\n",
      "Iter 158720, Minibatch Loss= 172.031540, Training Accuracy= 0.96875\n",
      "Iter 160000, Minibatch Loss= 46.291718, Training Accuracy= 0.97656\n",
      "Iter 161280, Minibatch Loss= 193.612808, Training Accuracy= 0.97656\n",
      "Iter 162560, Minibatch Loss= 194.584229, Training Accuracy= 0.96875\n",
      "Iter 163840, Minibatch Loss= 165.364227, Training Accuracy= 0.97656\n",
      "Iter 165120, Minibatch Loss= 194.127808, Training Accuracy= 0.97656\n",
      "Iter 166400, Minibatch Loss= 42.612534, Training Accuracy= 0.98438\n",
      "Iter 167680, Minibatch Loss= 148.357239, Training Accuracy= 0.97656\n",
      "Iter 168960, Minibatch Loss= 81.666550, Training Accuracy= 0.98438\n",
      "Iter 170240, Minibatch Loss= 59.235317, Training Accuracy= 0.96875\n",
      "Iter 171520, Minibatch Loss= 129.971375, Training Accuracy= 0.97656\n",
      "Iter 172800, Minibatch Loss= 88.927139, Training Accuracy= 0.98438\n",
      "Iter 174080, Minibatch Loss= 110.480606, Training Accuracy= 0.97656\n",
      "Iter 175360, Minibatch Loss= 162.427460, Training Accuracy= 0.96875\n",
      "Iter 176640, Minibatch Loss= 125.514702, Training Accuracy= 0.99219\n",
      "Iter 177920, Minibatch Loss= 47.490128, Training Accuracy= 0.98438\n",
      "Iter 179200, Minibatch Loss= 171.219360, Training Accuracy= 0.98438\n",
      "Iter 180480, Minibatch Loss= 160.448395, Training Accuracy= 0.97656\n",
      "Iter 181760, Minibatch Loss= 15.158218, Training Accuracy= 0.99219\n",
      "Iter 183040, Minibatch Loss= 186.794174, Training Accuracy= 0.97656\n",
      "Iter 184320, Minibatch Loss= 627.208862, Training Accuracy= 0.93750\n",
      "Iter 185600, Minibatch Loss= 81.770210, Training Accuracy= 0.97656\n",
      "Iter 186880, Minibatch Loss= 134.788162, Training Accuracy= 0.95312\n",
      "Iter 188160, Minibatch Loss= 197.437866, Training Accuracy= 0.97656\n",
      "Iter 189440, Minibatch Loss= 2.786758, Training Accuracy= 0.99219\n",
      "Iter 190720, Minibatch Loss= 429.330444, Training Accuracy= 0.96094\n",
      "Iter 192000, Minibatch Loss= 381.485474, Training Accuracy= 0.95312\n",
      "Iter 193280, Minibatch Loss= 321.565918, Training Accuracy= 0.96094\n",
      "Iter 194560, Minibatch Loss= 357.440094, Training Accuracy= 0.96094\n",
      "Iter 195840, Minibatch Loss= 263.318970, Training Accuracy= 0.95312\n",
      "Iter 197120, Minibatch Loss= 126.319519, Training Accuracy= 0.97656\n",
      "Iter 198400, Minibatch Loss= 135.008011, Training Accuracy= 0.98438\n",
      "Iter 199680, Minibatch Loss= 57.222519, Training Accuracy= 0.98438\n",
      "Optimization Finished!\n",
      "Testing Accuracy: 0.972656\n"
     ]
    }
   ],
   "source": [
    "# Launch the graph\n",
    "with tf.Session() as sess:\n",
    "    sess.run(init)\n",
    "    step = 1\n",
    "    # Keep training until reach max iterations\n",
    "    while step * batch_size < training_iters:\n",
    "        batch_x, batch_y = mnist.train.next_batch(batch_size)\n",
    "        # Run optimization op (backprop)\n",
    "        sess.run(optimizer, feed_dict={x: batch_x, y: batch_y,\n",
    "                                       keep_prob: dropout})\n",
    "        if step % display_step == 0:\n",
    "            # Calculate batch loss and accuracy\n",
    "            loss, acc = sess.run([cost, accuracy], feed_dict={x: batch_x,\n",
    "                                                              y: batch_y,\n",
    "                                                              keep_prob: 1.})\n",
    "            print (\"Iter \" + str(step*batch_size) + \", Minibatch Loss= \" + \\\n",
    "                  \"{:.6f}\".format(loss) + \", Training Accuracy= \" + \\\n",
    "                  \"{:.5f}\".format(acc))\n",
    "        step += 1\n",
    "    print (\"Optimization Finished!\")\n",
    "\n",
    "    # Calculate accuracy for 256 mnist test images\n",
    "    print (\"Testing Accuracy:\", \\\n",
    "        sess.run(accuracy, feed_dict={x: mnist.test.images[:256],\n",
    "                                      y: mnist.test.labels[:256],\n",
    "                                      keep_prob: 1.}))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [default]",
   "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": 1
}