MNIST_FCNN.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "# Fully Convolutional MNIST\n",
    "This is to get a feel for building classifiers in TF\n",
    "\n",
    "This model is fully convolutional, no fully connected layers\n",
    "\n",
    "It works okay, but is really weird about more than 2 layer, sometimes it works, others it doesn't\n",
    "\n",
    "As with any deepish NN, initialization is key. I did a blind manual search to find the values for the weight range and learning rate.\n",
    "\n",
    "With incorrect hyper parameters, the model seems to 'forget' some of the possible outputs and simply never predicts them."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import tensorflow as tf\n",
    "%matplotlib inline\n",
    "plt.rcParams['figure.figsize'] = (15,10)\n",
    "\n",
    "# auto reloading of modules\n",
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "# setup logging\n",
    "tf.logging.set_verbosity(tf.logging.INFO)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "# Settings\n",
    "batch_size = 1000\n",
    "epochs = 20 # number of epochs to train for\n",
    "learning_rate = .01\n",
    "momentum = .9\n",
    "print_every = 25 # print info about training at iterations\n",
    "test_every = 50 # number of iterations to go between doing 1 epoch of testing\n",
    "\n",
    "reg = 0.01 # regularization strength\n",
    "\n",
    "num_train = 60000 # 60000 max\n",
    "num_test = 5000 # 10000 max\n",
    "\n",
    "filter_depth = [10,10,10,10,10] # how many layers are in each convolution\n",
    "filter_extent = [5,5,5,5,5] # how big the filters are on each layer\n",
    "filter_stride = [1,1,1,1,1]\n",
    "pool_use = False # use max pooling?\n",
    "pool_stride = [2,2,2,2,2] # how big the stride is at every pool\n",
    "pool_extent = [2,2,2,2,2] # how wide to max pool over\n",
    "layer_padding = 'VALID'\n",
    "checkpoint_file = 'models/MNIST_FCNN.ckpt'\n",
    "\n",
    "img_shape = (28,28,1)\n",
    "\n",
    "log_file = 'logs/' # where to save the tf log files\n",
    "\n",
    "SEED = 0 # seed for random numbers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In conv and VALID\n",
      "After convolution 1 and pooling, width is 24.0\n",
      "In conv and VALID\n",
      "After convolution 2 and pooling, width is 20.0\n",
      "In conv and VALID\n",
      "After convolution 3 and pooling, width is 16.0\n",
      "In conv and VALID\n",
      "After convolution 4 and pooling, width is 12.0\n",
      "In conv and VALID\n",
      "After convolution 5 and pooling, width is 8.0\n",
      "Adding a 8x8 filter to remove spatial extents\n"
     ]
    }
   ],
   "source": [
    "# these are updates/modifications/ utilites based on settings, do not change\n",
    "\n",
    "# fixed seed for consistancy\n",
    "np.random.seed(SEED)\n",
    "\n",
    "# set up number of batches, make sure its a valid number\n",
    "num_batches_train = num_train/batch_size\n",
    "num_batches_test = num_test/batch_size\n",
    "assert num_train%batch_size == 0, 'The number of training samples must be divisible by the batch size'\n",
    "assert num_test%batch_size == 0, 'The number of test samples must be divisible by the batch size'\n",
    "\n",
    "# determine how big the filter must be on the last layer to get a [nx1x10] vector\n",
    "layer_width = float(img_shape[0])\n",
    "counter = 0\n",
    "for FE,FS,PS,PE in zip(filter_extent, filter_stride, pool_stride, pool_extent):\n",
    "    counter += 1\n",
    "    \n",
    "    if layer_padding == 'VALID':\n",
    "        print('In conv and VALID')\n",
    "        layer_width = np.ceil(float(layer_width - FE + 1)/float(FS))\n",
    "    elif layer_padding == 'SAME':\n",
    "        print('In conv and SAME')\n",
    "        layer_width = np.ceil(float(layer_width)/float(FS))\n",
    "    else:\n",
    "        print('\\n\\n--ERROR--\\nInvalid padding scheme\\n--ERROR--\\n\\n')\n",
    "        break\n",
    "    \n",
    "    if pool_use:\n",
    "        if layer_padding == 'VALID':\n",
    "            print('In pooling and VALID')\n",
    "            layer_width = np.ceil(float(layer_width - PE + 1)/float(PS))\n",
    "        elif layer_padding == 'SAME':\n",
    "            print('In pooling and SAME')\n",
    "            layer_width = np.ceil(float(layer_width)/float(PS))\n",
    "        else:\n",
    "            print('\\n\\n--ERROR--\\nInvalid padding scheme\\n--ERROR--\\n\\n')\n",
    "            break\n",
    "        \n",
    "    assert layer_width %int(layer_width) == 0, 'Current filter sizes, padding, stride, and pooling cause a non-integer output of layer {}'.format(counter)\n",
    "    assert layer_width >= 1.0, 'Current filter sizes, padding, stride, and pooling cause a negative output of layer {}'.format(counter)\n",
    "    print('After convolution {} and pooling, width is {}'.format(counter,layer_width))\n",
    "print('Adding a {:.0f}x{:.0f} filter to remove spatial extents'.format(layer_width,layer_width))\n",
    "filter_extent.append(layer_width)\n",
    "filter_depth.append(10) # 10 outputs\n",
    "filter_stride.append(1) # arbitrary, it can only do it once anyway"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "## Using code from 3_mnist_from_scratch tutorial to deal with downloads and working with MNIST data\n",
    "\n",
    "Took the code and made a simple module to make everything cleaner"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "('Already downloaded', 'train-images-idx3-ubyte.gz')\n",
      "('Already downloaded', 'train-labels-idx1-ubyte.gz')\n",
      "('Already downloaded', 't10k-images-idx3-ubyte.gz')\n",
      "('Already downloaded', 't10k-labels-idx1-ubyte.gz')\n",
      "('Extracting', '/tmp/mnist-data/train-images-idx3-ubyte.gz')\n",
      "('Extracting', '/tmp/mnist-data/t10k-images-idx3-ubyte.gz')\n",
      "('Extracting', '/tmp/mnist-data/train-labels-idx1-ubyte.gz')\n",
      "('Extracting', '/tmp/mnist-data/t10k-labels-idx1-ubyte.gz')\n"
     ]
    }
   ],
   "source": [
    "import getMNIST\n",
    "fNames = getMNIST.downloadMNIST()\n",
    "\n",
    "train_data = getMNIST.extract_data(fNames[0],num_train)\n",
    "test_data = getMNIST.extract_data(fNames[2],num_test)\n",
    "\n",
    "train_labels = getMNIST.extract_labels(fNames[1], num_train)\n",
    "test_labels = getMNIST.extract_labels(fNames[3], num_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "## Building the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "# Set up inputs and outputs\n",
    "X = tf.placeholder(tf.float32, shape = [batch_size,img_shape[0],img_shape[1],img_shape[2]],name='Input')\n",
    "Y = tf.placeholder(tf.float32, shape = [batch_size,10],name = 'Labels') # 10 labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "conv3: <tf.Variable 'conv_weights_3:0' shape=(5, 5, 10, 10) dtype=float32_ref>\n",
      "conv3: <tf.Variable 'conv_bias_3:0' shape=(10,) dtype=float32_ref>\n",
      "conv2: <tf.Variable 'conv_weights_2:0' shape=(5, 5, 10, 10) dtype=float32_ref>\n",
      "conv2: <tf.Variable 'conv_bias_2:0' shape=(10,) dtype=float32_ref>\n",
      "conv1: <tf.Variable 'conv_weights_1:0' shape=(5, 5, 1, 10) dtype=float32_ref>\n",
      "conv1: <tf.Variable 'conv_bias_1:0' shape=(10,) dtype=float32_ref>\n",
      "conv6: <tf.Variable 'conv_weights_6:0' shape=(8, 8, 10, 10) dtype=float32_ref>\n",
      "conv6: <tf.Variable 'conv_bias_6:0' shape=(10,) dtype=float32_ref>\n",
      "conv5: <tf.Variable 'conv_weights_5:0' shape=(5, 5, 10, 10) dtype=float32_ref>\n",
      "conv5: <tf.Variable 'conv_bias_5:0' shape=(10,) dtype=float32_ref>\n",
      "conv4: <tf.Variable 'conv_weights_4:0' shape=(5, 5, 10, 10) dtype=float32_ref>\n",
      "conv4: <tf.Variable 'conv_bias_4:0' shape=(10,) dtype=float32_ref>\n"
     ]
    }
   ],
   "source": [
    "conv_weights = {}\n",
    "conv_bias = {}\n",
    "\n",
    "prev_layer_depth = img_shape[2]\n",
    "for layer in range(len(filter_depth)):\n",
    "    layer_name = 'conv'+str(layer+1)\n",
    "    \n",
    "    filter_shape = [int(filter_extent[layer]),int(filter_extent[layer]), prev_layer_depth, int(filter_depth[layer])]\n",
    "    \n",
    "    \n",
    "    # set up the scales for the weights\n",
    "    # see https://www.tensorflow.org/versions/r0.11/api_docs/python/contrib.layers/initializers#xavier_initializer \n",
    "    # for more info, specifically varianc_scaling_initializer and http://arxiv.org/pdf/1502.01852v1.pdf\n",
    "    fan_in = prev_layer_depth*filter_extent[layer]**2\n",
    "    weight_scale = np.sqrt(2./fan_in)\n",
    "    \n",
    "    conv_weights[layer_name] = tf.Variable(tf.truncated_normal(filter_shape,\n",
    "                                              stddev = weight_scale,\n",
    "                                              seed = SEED),\n",
    "                                          name = 'conv_weights_' + str(layer+1))\n",
    "    \n",
    "    conv_bias[layer_name] = tf.Variable(tf.zeros([int(filter_depth[layer])],dtype=tf.float32),\n",
    "                                       name = 'conv_bias_' + str(layer+1))    \n",
    "    \n",
    "    prev_layer_depth = filter_depth[layer]\n",
    "    \n",
    "for layer in conv_weights:\n",
    "    print('{}: {}'.format(layer, conv_weights[layer]))\n",
    "    print('{}: {}'.format(layer, conv_bias[layer]))\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "640.0"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "fan_in"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.055901699437494741"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "weight_scale"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "conv1\n",
      "conv2\n",
      "conv3\n",
      "conv4\n",
      "conv5\n",
      "conv6\n"
     ]
    }
   ],
   "source": [
    "# build the model\n",
    "# dicts to store intermediate values\n",
    "\n",
    "conv = {}\n",
    "relu = {}\n",
    "pool = {}\n",
    "previous_layer = X\n",
    "\n",
    "for layer in range(len(filter_depth)):\n",
    "    layer_name = 'conv'+str(layer+1)\n",
    "    print layer_name\n",
    "    conv[layer_name] = tf.nn.conv2d(previous_layer,\n",
    "                                   conv_weights[layer_name],\n",
    "                                   strides=[1,filter_stride[layer],filter_stride[layer],1],\n",
    "                                   padding='VALID',\n",
    "                                   name = 'conv_' + str(layer+1))\n",
    "    \n",
    "    biased = tf.nn.bias_add(conv[layer_name], \n",
    "                            conv_bias[layer_name],\n",
    "                            name = 'add_bias_' + str(layer+1))\n",
    "    #apply relu\n",
    "    relu[layer_name] = tf.nn.relu(biased,\n",
    "                                  name = 'relu_' + str(layer+1))\n",
    "    \n",
    "    # check if doing pooling and if it is the last layer\n",
    "    # no pooling on the last layer\n",
    "    if pool_use and not layer == len(filter_depth)-1:\n",
    "        print layer\n",
    "        pool[layer_name] = tf.nn.max_pool(relu[layer_name],\n",
    "                                         ksize=[1,2,2,1],\n",
    "                                         strides = [1,2,2,1],\n",
    "                                         padding = 'VALID',\n",
    "                                         name = 'pool_' + str(layer+1))\n",
    "        previous_layer = pool[layer_name]\n",
    "    else: # not the last layer and not pooling \n",
    "        previous_layer = relu[layer_name]\n",
    "\n",
    "    \n",
    "loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=Y,logits = previous_layer, name = 'softmax'),\n",
    "                     name = 'loss')\n",
    "\n",
    "# regularization loss\n",
    "for ii in conv_weights:\n",
    "    loss = tf.add(loss,reg * tf.nn.l2_loss(conv_weights[ii],\n",
    "                                           name = 'Regularization_Loss_Addition'),\n",
    "                      name = 'reg_loss_summing')\n",
    "    \n",
    "# get rid of singular dims in tensor\n",
    "scores = tf.squeeze(previous_layer, name = 'prediction_squeeze')\n",
    "predictions = tf.nn.softmax(logits = scores,\n",
    "                           name = 'predictions_softmax')\n",
    "\n",
    "num_correct = tf.reduce_sum(tf.cast(tf.equal(tf.argmax(predictions,1), tf.argmax(Y,1)),tf.float32),\n",
    "                           name = 'num_correct')\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.055901699437494741"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "weight_scale"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<tf.Variable 'conv_weights_1:0' shape=(5, 5, 1, 10) dtype=float32_ref>,\n",
       " <tf.Variable 'conv_bias_1:0' shape=(10,) dtype=float32_ref>,\n",
       " <tf.Variable 'conv_weights_2:0' shape=(5, 5, 10, 10) dtype=float32_ref>,\n",
       " <tf.Variable 'conv_bias_2:0' shape=(10,) dtype=float32_ref>,\n",
       " <tf.Variable 'conv_weights_3:0' shape=(5, 5, 10, 10) dtype=float32_ref>,\n",
       " <tf.Variable 'conv_bias_3:0' shape=(10,) dtype=float32_ref>,\n",
       " <tf.Variable 'conv_weights_4:0' shape=(5, 5, 10, 10) dtype=float32_ref>,\n",
       " <tf.Variable 'conv_bias_4:0' shape=(10,) dtype=float32_ref>,\n",
       " <tf.Variable 'conv_weights_5:0' shape=(5, 5, 10, 10) dtype=float32_ref>,\n",
       " <tf.Variable 'conv_bias_5:0' shape=(10,) dtype=float32_ref>,\n",
       " <tf.Variable 'conv_weights_6:0' shape=(8, 8, 10, 10) dtype=float32_ref>,\n",
       " <tf.Variable 'conv_bias_6:0' shape=(10,) dtype=float32_ref>]"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tf.trainable_variables()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "optimizer = tf.train.MomentumOptimizer(learning_rate, momentum).minimize(loss)\n",
    "init = tf.global_variables_initializer()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "with tf.Session() as sess:\n",
    "    sess.run(tf.global_variables_initializer())\n",
    "    \n",
    "    batch = 0\n",
    "    batch_start = batch*batch_size\n",
    "    batch_end = batch_start + batch_size\n",
    "    \n",
    "    # run it\n",
    "    out = sess.run(loss,\n",
    "                   feed_dict={X:train_data[batch_start:batch_end,:],\n",
    "                              Y:train_labels[batch_start:batch_end]})\n",
    "    merged = tf.summary.merge_all()\n",
    "    file_writer = tf.summary.FileWriter(log_file, sess.graph)\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration: 0/1200, batch 0 on epoch 0, \tLoss: 2.79e+00, Train Acc: 0.072\n",
      "Testing Loss: 1.40e+01\tTesting Accuracy: 0.079\n",
      "\n",
      "Iteration: 25/1200, batch 25 on epoch 0, \tLoss: 2.23e+00, Train Acc: 0.467\n",
      "Iteration: 50/1200, batch 50 on epoch 0, \tLoss: 1.27e+00, Train Acc: 0.750\n",
      "Testing Loss: 6.57e+00\tTesting Accuracy: 0.744\n",
      "\n",
      "Iteration: 75/1200, batch 15 on epoch 1, \tLoss: 9.99e-01, Train Acc: 0.826\n",
      "Iteration: 100/1200, batch 40 on epoch 1, \tLoss: 6.71e-01, Train Acc: 0.927\n",
      "Testing Loss: 3.84e+00\tTesting Accuracy: 0.899\n",
      "\n",
      "Iteration: 125/1200, batch 5 on epoch 2, \tLoss: 6.63e-01, Train Acc: 0.930\n",
      "Iteration: 150/1200, batch 30 on epoch 2, \tLoss: 6.24e-01, Train Acc: 0.936\n",
      "Testing Loss: 3.25e+00\tTesting Accuracy: 0.925\n",
      "\n",
      "Iteration: 175/1200, batch 55 on epoch 2, \tLoss: 5.38e-01, Train Acc: 0.963\n",
      "Iteration: 200/1200, batch 20 on epoch 3, \tLoss: 5.67e-01, Train Acc: 0.945\n",
      "Testing Loss: 2.84e+00\tTesting Accuracy: 0.942\n",
      "\n",
      "Iteration: 225/1200, batch 45 on epoch 3, \tLoss: 5.23e-01, Train Acc: 0.942\n",
      "Iteration: 250/1200, batch 10 on epoch 4, \tLoss: 4.66e-01, Train Acc: 0.968\n",
      "Testing Loss: 2.59e+00\tTesting Accuracy: 0.948\n",
      "\n",
      "Iteration: 275/1200, batch 35 on epoch 4, \tLoss: 4.53e-01, Train Acc: 0.965\n",
      "Iteration: 300/1200, batch 0 on epoch 5, \tLoss: 4.67e-01, Train Acc: 0.963\n",
      "Testing Loss: 2.37e+00\tTesting Accuracy: 0.953\n",
      "\n",
      "Iteration: 325/1200, batch 25 on epoch 5, \tLoss: 4.18e-01, Train Acc: 0.967\n",
      "Iteration: 350/1200, batch 50 on epoch 5, \tLoss: 4.28e-01, Train Acc: 0.964\n",
      "Testing Loss: 2.18e+00\tTesting Accuracy: 0.958\n",
      "\n",
      "Iteration: 375/1200, batch 15 on epoch 6, \tLoss: 4.01e-01, Train Acc: 0.972\n",
      "Iteration: 400/1200, batch 40 on epoch 6, \tLoss: 3.65e-01, Train Acc: 0.983\n",
      "Testing Loss: 2.07e+00\tTesting Accuracy: 0.957\n",
      "\n",
      "Iteration: 425/1200, batch 5 on epoch 7, \tLoss: 3.79e-01, Train Acc: 0.972\n",
      "Iteration: 450/1200, batch 30 on epoch 7, \tLoss: 3.67e-01, Train Acc: 0.970\n",
      "Testing Loss: 2.00e+00\tTesting Accuracy: 0.958\n",
      "\n",
      "Iteration: 475/1200, batch 55 on epoch 7, \tLoss: 3.42e-01, Train Acc: 0.972\n",
      "Iteration: 500/1200, batch 20 on epoch 8, \tLoss: 3.65e-01, Train Acc: 0.969\n",
      "Testing Loss: 1.81e+00\tTesting Accuracy: 0.963\n",
      "\n",
      "Iteration: 525/1200, batch 45 on epoch 8, \tLoss: 3.44e-01, Train Acc: 0.958\n",
      "Iteration: 550/1200, batch 10 on epoch 9, \tLoss: 3.22e-01, Train Acc: 0.971\n",
      "Testing Loss: 1.76e+00\tTesting Accuracy: 0.962\n",
      "\n",
      "Iteration: 575/1200, batch 35 on epoch 9, \tLoss: 3.26e-01, Train Acc: 0.967\n",
      "Iteration: 600/1200, batch 0 on epoch 10, \tLoss: 3.24e-01, Train Acc: 0.975\n",
      "Testing Loss: 1.60e+00\tTesting Accuracy: 0.965\n",
      "\n",
      "Iteration: 625/1200, batch 25 on epoch 10, \tLoss: 2.88e-01, Train Acc: 0.976\n",
      "Iteration: 650/1200, batch 50 on epoch 10, \tLoss: 2.88e-01, Train Acc: 0.976\n",
      "Testing Loss: 1.55e+00\tTesting Accuracy: 0.965\n",
      "\n",
      "Iteration: 675/1200, batch 15 on epoch 11, \tLoss: 2.90e-01, Train Acc: 0.977\n",
      "Iteration: 700/1200, batch 40 on epoch 11, \tLoss: 2.59e-01, Train Acc: 0.979\n",
      "Testing Loss: 1.47e+00\tTesting Accuracy: 0.967\n",
      "\n",
      "Iteration: 725/1200, batch 5 on epoch 12, \tLoss: 2.82e-01, Train Acc: 0.975\n",
      "Iteration: 750/1200, batch 30 on epoch 12, \tLoss: 2.62e-01, Train Acc: 0.974\n",
      "Testing Loss: 1.60e+00\tTesting Accuracy: 0.957\n",
      "\n",
      "Iteration: 775/1200, batch 55 on epoch 12, \tLoss: 2.58e-01, Train Acc: 0.976\n",
      "Iteration: 800/1200, batch 20 on epoch 13, \tLoss: 2.77e-01, Train Acc: 0.972\n",
      "Testing Loss: 1.38e+00\tTesting Accuracy: 0.965\n",
      "\n",
      "Iteration: 825/1200, batch 45 on epoch 13, \tLoss: 2.48e-01, Train Acc: 0.977\n",
      "Iteration: 850/1200, batch 10 on epoch 14, \tLoss: 2.32e-01, Train Acc: 0.979\n",
      "Testing Loss: 1.34e+00\tTesting Accuracy: 0.966\n",
      "\n",
      "Iteration: 875/1200, batch 35 on epoch 14, \tLoss: 2.34e-01, Train Acc: 0.980\n",
      "Iteration: 900/1200, batch 0 on epoch 15, \tLoss: 2.63e-01, Train Acc: 0.973\n",
      "Testing Loss: 1.29e+00\tTesting Accuracy: 0.967\n",
      "\n",
      "Iteration: 925/1200, batch 25 on epoch 15, \tLoss: 2.28e-01, Train Acc: 0.978\n",
      "Iteration: 950/1200, batch 50 on epoch 15, \tLoss: 2.20e-01, Train Acc: 0.982\n",
      "Testing Loss: 1.18e+00\tTesting Accuracy: 0.968\n",
      "\n",
      "Iteration: 975/1200, batch 15 on epoch 16, \tLoss: 2.19e-01, Train Acc: 0.979\n",
      "Iteration: 1000/1200, batch 40 on epoch 16, \tLoss: 1.98e-01, Train Acc: 0.982\n",
      "Testing Loss: 1.17e+00\tTesting Accuracy: 0.969\n",
      "\n",
      "Iteration: 1025/1200, batch 5 on epoch 17, \tLoss: 2.33e-01, Train Acc: 0.978\n",
      "Iteration: 1050/1200, batch 30 on epoch 17, \tLoss: 2.01e-01, Train Acc: 0.975\n",
      "Testing Loss: 1.13e+00\tTesting Accuracy: 0.969\n",
      "\n",
      "Iteration: 1075/1200, batch 55 on epoch 17, \tLoss: 1.97e-01, Train Acc: 0.978\n",
      "Iteration: 1100/1200, batch 20 on epoch 18, \tLoss: 2.26e-01, Train Acc: 0.968\n",
      "Testing Loss: 1.10e+00\tTesting Accuracy: 0.969\n",
      "\n",
      "Iteration: 1125/1200, batch 45 on epoch 18, \tLoss: 2.00e-01, Train Acc: 0.979\n",
      "Iteration: 1150/1200, batch 10 on epoch 19, \tLoss: 1.88e-01, Train Acc: 0.986\n",
      "Testing Loss: 1.10e+00\tTesting Accuracy: 0.968\n",
      "\n",
      "Iteration: 1175/1200, batch 35 on epoch 19, \tLoss: 1.83e-01, Train Acc: 0.981\n",
      "Model saved in file: models/MNIST_FCNN.ckpt\n"
     ]
    }
   ],
   "source": [
    "# loss history\n",
    "losses_train = np.array([])\n",
    "losses_test = np.array([])\n",
    "# accuracy history\n",
    "acc_train = np.array([])\n",
    "acc_test = np.array([])\n",
    "\n",
    "saver = tf.train.Saver()\n",
    "\n",
    "iteration = -1 # counter\n",
    "with tf.Session() as sess:\n",
    "    sess.run(tf.global_variables_initializer())\n",
    "    \n",
    "    for epoch in xrange(epochs):        \n",
    "        for batch in xrange(num_batches_train):\n",
    "            iteration += 1\n",
    "            \n",
    "            batch_start = batch*batch_size\n",
    "            batch_end = batch_start + batch_size\n",
    "            \n",
    "            # We want different info based on the iteration. \n",
    "            # We dont actually need to compute everything on every iteration            \n",
    "            if iteration % print_every == 0:\n",
    "                solveFor = [optimizer, loss, predictions, num_correct]\n",
    "            else:\n",
    "                solveFor = [optimizer, loss]\n",
    "            \n",
    "            # run it\n",
    "            out = sess.run(solveFor,\n",
    "                           feed_dict={X:train_data[batch_start:batch_end,:],\n",
    "                                      Y:train_labels[batch_start:batch_end]})\n",
    "            \n",
    "            if iteration % print_every == 0:\n",
    "                _, l, pred, correct = out\n",
    "                acc_train = np.append(acc_train, correct.astype(np.float32)/float(batch_size))\n",
    "            else:\n",
    "                _, l = out\n",
    "                \n",
    "            losses_train = np.append(losses_train,l)\n",
    "            \n",
    "            if iteration % print_every == 0:\n",
    "                print('Iteration: {}/{}, batch {} on epoch {}, \\tLoss: {:.2e}, Train Acc: {:.3f}'.format(iteration, \n",
    "                                                                                            epochs*num_batches_train,\n",
    "                                                                                            batch, \n",
    "                                                                                            epoch, \n",
    "                                                                                            l, \n",
    "                                                                                            acc_train[-1]))\n",
    "            \n",
    "            if iteration % test_every == 0:\n",
    "                l_test = 0\n",
    "                correct_test = 0\n",
    "                for test_it in xrange(num_batches_test):\n",
    "                    batch_start = test_it*batch_size\n",
    "                    batch_end = batch_start + batch_size\n",
    "                    l, correct = sess.run([loss,num_correct],\n",
    "                                         feed_dict={X:test_data[batch_start:batch_end,:],\n",
    "                                                    Y:test_labels[batch_start:batch_end]})\n",
    "                    correct_test += correct\n",
    "                    l_test += l\n",
    "                losses_test = np.append(losses_test,l_test.astype(np.float32)/float(num_batches_test))\n",
    "                acc_test = np.append(acc_test, correct_test.astype(np.float32)/float(num_test))\n",
    "                print('Testing Loss: {:.2e}\\tTesting Accuracy: {:.3f}\\n'.format(l_test, acc_test[-1]))\n",
    "    save_path = saver.save(sess, checkpoint_file)\n",
    "    print(\"Model saved in file: %s\" % save_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f783ff60390>"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA3sAAAGDCAYAAACSkwm+AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7858ce29d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize the cost and log cost from training\n",
    "plt.rcParams['figure.figsize'] = (15,6)\n",
    "\n",
    "plt.subplot(121)\n",
    "plt.plot(np.linspace(0,iteration,losses_train.shape[0]),losses_train,'b',label='Training Loss')\n",
    "plt.plot(np.linspace(0,iteration,losses_test.shape[0]),losses_test,'r',label='Testing Loss')\n",
    "plt.title('Cost')\n",
    "plt.xlabel('Iteration Number')\n",
    "plt.ylabel('Cost')\n",
    "plt.legend()\n",
    "\n",
    "plt.subplot(122)\n",
    "plt.plot(np.log(np.linspace(1,iteration,losses_train.shape[0])),np.log(losses_train),'b',label='Training Loss')\n",
    "plt.plot(np.log(np.linspace(1,iteration,losses_test.shape[0])),np.log(losses_test),'r',label='Testing Loss')\n",
    "plt.title('Log Cost')\n",
    "plt.xlabel('log(Iteration Number)')\n",
    "plt.ylabel('log(Cost)')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f783415dd10>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA3sAAAEWCAYAAADfOI5OAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nERERERERNfRfqM8hiLckDhQAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7858ce2d50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# See how the accuracy plays out between the test and training set\n",
    "plt.rcParams['figure.figsize'] = (15,4)\n",
    "\n",
    "plt.figure()\n",
    "plt.plot(np.linspace(0,iteration,acc_train.shape[0]),acc_train,label = 'Train Accuracy')\n",
    "plt.plot(np.linspace(0,iteration,acc_test.shape[0]),acc_test,label = 'Test Accuracy')\n",
    "plt.title('Accuracy')\n",
    "plt.xlabel('Iteration Number')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "## Visualizing the model outputs\n",
    "Want to see how the model did, specifically what is was confused on"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Restoring parameters from models/MNIST_FCNN.ckpt\n",
      "Finished loading the model\n",
      "Finished predictions for training data\n",
      "Finished predictions for testing data\n",
      "Done!\n"
     ]
    }
   ],
   "source": [
    "# Generate predictions for the test and train set\n",
    "saver = tf.train.Saver()\n",
    "\n",
    "pred_train = np.zeros((num_train,10))\n",
    "pred_test = np.zeros((num_test,10))\n",
    "\n",
    "\n",
    "with tf.Session() as sess:\n",
    "    # load the model from before\n",
    "    saver.restore(sess, checkpoint_file)\n",
    "    print('Finished loading the model')\n",
    "    \n",
    "    # batch over the entire training set and record predictions\n",
    "    for it in xrange(num_batches_train):\n",
    "        batch_start = it*batch_size\n",
    "        batch_end = batch_start + batch_size\n",
    "        pred_train[batch_start:batch_end,:] = sess.run(predictions,\n",
    "                                                     feed_dict={X:train_data[batch_start:batch_end,:],\n",
    "                                                                Y:train_labels[batch_start:batch_end]})\n",
    "    print('Finished predictions for training data')\n",
    "    \n",
    "    # batch over the entire testing set and record predictions\n",
    "    for it in xrange(num_batches_test):\n",
    "        batch_start = it*batch_size\n",
    "        batch_end = batch_start + batch_size\n",
    "        pred_test[batch_start:batch_end,:] = sess.run(predictions,\n",
    "                                                     feed_dict={X:test_data[batch_start:batch_end,:],\n",
    "                                                                Y:test_labels[batch_start:batch_end]})\n",
    "    print('Finished predictions for testing data')\n",
    "print('Done!')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "def error_rate(predictions, labels):\n",
    "    \"\"\"Return the error rate and confusions.\"\"\"    \n",
    "    actual = np.argmax(labels,1)\n",
    "    pred = np.argmax(predictions,1)\n",
    "    correct = np.sum(pred == actual)\n",
    "    total = pred.shape[0]\n",
    "    error = 100.0 - (100 * float(correct) / float(total))\n",
    "\n",
    "    confusions = np.zeros([10, 10], np.float32)\n",
    "    for ii in xrange(predictions.shape[0]):        \n",
    "        confusions[pred[ii], actual[ii]] += 1\n",
    "    \n",
    "    return error, confusions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train error: 2.2%\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAArkAAAK9CAYAAAAgz3SiAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f78340f9350>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Make confusion plot of the training set\n",
    "plt.rcParams['figure.figsize'] = (20, 12)\n",
    "\n",
    "error_train, confusions_train = error_rate(pred_train, train_labels)\n",
    "print('Train error: %.1f%%' % error_train)\n",
    "\n",
    "plt.xlabel('Actual')\n",
    "plt.ylabel('Predicted')\n",
    "plt.grid(False)\n",
    "plt.xticks(np.arange(10))\n",
    "plt.yticks(np.arange(10))\n",
    "plt.imshow(confusions_train, cmap=plt.cm.jet, interpolation='nearest');\n",
    "\n",
    "for i, cas in enumerate(confusions_train):\n",
    "    for j, count in enumerate(cas):\n",
    "        if count > 0:\n",
    "            xoff = .07 * len(str(count))\n",
    "            plt.text(j-xoff, i+.2, int(count), fontsize=12, color='white')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test error: 3.0%\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAArkAAAK9CAYAAAAgz3SiAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nciVJklQ7llxJkiTVzv8BcTw9LEQQfqIAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7834196a50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Make confusion plot of the test set\n",
    "plt.rcParams['figure.figsize'] = (20, 12)\n",
    "\n",
    "error_test, confusions_test = error_rate(pred_test, test_labels)\n",
    "print('Test error: %.1f%%' % error_test)\n",
    "\n",
    "plt.xlabel('Actual')\n",
    "plt.ylabel('Predicted')\n",
    "plt.grid(False)\n",
    "plt.xticks(np.arange(10))\n",
    "plt.yticks(np.arange(10))\n",
    "plt.imshow(confusions_train, cmap=plt.cm.jet, interpolation='nearest');\n",
    "\n",
    "for i, cas in enumerate(confusions_test):\n",
    "    for j, count in enumerate(cas):\n",
    "        if count > 0:\n",
    "            xoff = .07 * len(str(count))\n",
    "            plt.text(j-xoff, i+.2, int(count), fontsize=12, color='white')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}