4_convolutions.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "4embtkV0pNxM"
   },
   "source": [
    "Deep Learning\n",
    "=============\n",
    "\n",
    "Assignment 4\n",
    "------------\n",
    "\n",
    "Previously in `2_fullyconnected.ipynb` and `3_regularization.ipynb`, we trained fully connected networks to classify [notMNIST](http://yaroslavvb.blogspot.com/2011/09/notmnist-dataset.html) characters.\n",
    "\n",
    "The goal of this assignment is make the neural network convolutional."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "cellView": "both",
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "tm2CQN_Cpwj0"
   },
   "outputs": [],
   "source": [
    "# These are all the modules we'll be using later. Make sure you can import them\n",
    "# before proceeding further.\n",
    "from __future__ import print_function\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from six.moves import cPickle as pickle\n",
    "from six.moves import range\n",
    "import math\n",
    "from IPython.display import clear_output\n",
    "\n",
    "from matplotlib import pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "cellView": "both",
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "output_extras": [
      {
       "item_id": 1
      }
     ]
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 11948,
     "status": "ok",
     "timestamp": 1446658914837,
     "user": {
      "color": "",
      "displayName": "",
      "isAnonymous": false,
      "isMe": true,
      "permissionId": "",
      "photoUrl": "",
      "sessionId": "0",
      "userId": ""
     },
     "user_tz": 480
    },
    "id": "y3-cj1bpmuxc",
    "outputId": "016b1a51-0290-4b08-efdb-8c95ffc3cd01"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set (200000, 28, 28) (200000,)\n",
      "Validation set (10000, 28, 28) (10000,)\n",
      "Test set (10000, 28, 28) (10000,)\n",
      "Deduplicated validation set (8093, 28, 28) (8093,)\n",
      "Deduplicated test set (7741, 28, 28) (7741,)\n"
     ]
    }
   ],
   "source": [
    "pickle_file = 'notMNIST.pickle'\n",
    "\n",
    "with open(pickle_file, 'rb') as f:\n",
    "  save = pickle.load(f)\n",
    "  train_dataset = save['train_dataset']\n",
    "  train_labels = save['train_labels']\n",
    "  valid_dataset = save['valid_dataset']\n",
    "  valid_labels = save['valid_labels']\n",
    "  test_dataset = save['test_dataset']\n",
    "  test_labels = save['test_labels']\n",
    "  dedup_valid_dataset = save['dedup_valid_dataset']\n",
    "  dedup_valid_labels = save['dedup_valid_labels']\n",
    "  dedup_test_dataset = save['dedup_test_dataset']\n",
    "  dedup_test_labels = save['dedup_test_labels']\n",
    "  del save  # hint to help gc free up memory\n",
    "  print('Training set', train_dataset.shape, train_labels.shape)\n",
    "  print('Validation set', valid_dataset.shape, valid_labels.shape)\n",
    "  print('Test set', test_dataset.shape, test_labels.shape)\n",
    "  print('Deduplicated validation set', dedup_valid_dataset.shape, dedup_valid_labels.shape)\n",
    "  print('Deduplicated test set', dedup_test_dataset.shape, dedup_test_labels.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "L7aHrm6nGDMB"
   },
   "source": [
    "Reformat into a TensorFlow-friendly shape:\n",
    "- convolutions need the image data formatted as a cube (width by height by #channels)\n",
    "- labels as float 1-hot encodings."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "cellView": "both",
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "output_extras": [
      {
       "item_id": 1
      }
     ]
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 11952,
     "status": "ok",
     "timestamp": 1446658914857,
     "user": {
      "color": "",
      "displayName": "",
      "isAnonymous": false,
      "isMe": true,
      "permissionId": "",
      "photoUrl": "",
      "sessionId": "0",
      "userId": ""
     },
     "user_tz": 480
    },
    "id": "IRSyYiIIGIzS",
    "outputId": "650a208c-8359-4852-f4f5-8bf10e80ef6c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set (200000, 28, 28, 1) (200000, 10)\n",
      "Validation set (10000, 28, 28, 1) (10000, 10)\n",
      "Test set (10000, 28, 28, 1) (10000, 10)\n",
      "Deduplicated validation set (8093, 28, 28, 1) (8093, 10)\n",
      "Deduplicated test set (7741, 28, 28, 1) (7741, 10)\n"
     ]
    }
   ],
   "source": [
    "image_size = 28\n",
    "num_labels = 10\n",
    "num_channels = 1 # grayscale\n",
    "\n",
    "import numpy as np\n",
    "\n",
    "def reformat(dataset, labels):\n",
    "  dataset = dataset.reshape(\n",
    "    (-1, image_size, image_size, num_channels)).astype(np.float32)\n",
    "  labels = (np.arange(num_labels) == labels[:,None]).astype(np.float32)\n",
    "  return dataset, labels\n",
    "\n",
    "train_dataset, train_labels = reformat(train_dataset, train_labels)\n",
    "orig_valid_dataset, orig_valid_labels = reformat(valid_dataset, valid_labels)\n",
    "orig_test_dataset, orig_test_labels = reformat(test_dataset, test_labels)\n",
    "dedup_valid_dataset, dedup_valid_labels = reformat(dedup_valid_dataset, dedup_valid_labels)\n",
    "dedup_test_dataset, dedup_test_labels = reformat(dedup_test_dataset, dedup_test_labels)\n",
    "\n",
    "print('Training set', train_dataset.shape, train_labels.shape)\n",
    "print('Validation set', orig_valid_dataset.shape, orig_valid_labels.shape)\n",
    "print('Test set', orig_test_dataset.shape, orig_test_labels.shape)\n",
    "print('Deduplicated validation set', dedup_valid_dataset.shape, dedup_valid_labels.shape)\n",
    "print('Deduplicated test set', dedup_test_dataset.shape, dedup_test_labels.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "valid_dataset = dedup_valid_dataset\n",
    "valid_labels = dedup_valid_labels\n",
    "test_dataset = dedup_test_dataset\n",
    "test_labels = dedup_test_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "valid_dataset = orig_valid_dataset\n",
    "valid_labels = orig_valid_labels\n",
    "test_dataset = orig_test_dataset\n",
    "test_labels = orig_test_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "cellView": "both",
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "AgQDIREv02p1"
   },
   "outputs": [],
   "source": [
    "def accuracy(predictions, labels):\n",
    "  return (100.0 * np.sum(np.equal(np.argmax(predictions, 1), np.argmax(labels, 1)))\n",
    "          / predictions.shape[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "5rhgjmROXu2O"
   },
   "source": [
    "Let's build a small network with two convolutional layers, followed by one fully connected layer. Convolutional networks are more expensive computationally, so we'll limit its depth and number of fully connected nodes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "cellView": "both",
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "IZYv70SvvOan"
   },
   "outputs": [],
   "source": [
    "batch_size = 16\n",
    "patch_size = 5\n",
    "depth = 16\n",
    "num_hidden = 64\n",
    "\n",
    "graph = tf.Graph()\n",
    "\n",
    "with graph.as_default(), tf.device('/cpu:0'):\n",
    "\n",
    "  # Input data.\n",
    "  tf_train_dataset = tf.placeholder(\n",
    "    tf.float32, shape=(batch_size, image_size, image_size, num_channels))\n",
    "  tf_train_labels = tf.placeholder(tf.float32, shape=(batch_size, num_labels))\n",
    "  tf_valid_dataset = tf.constant(valid_dataset)\n",
    "  tf_test_dataset = tf.constant(test_dataset)\n",
    "  \n",
    "  # Variables.\n",
    "  layer1_weights = tf.Variable(tf.truncated_normal(\n",
    "      [patch_size, patch_size, num_channels, depth], stddev=0.1))\n",
    "  layer1_biases = tf.Variable(tf.zeros([depth]))\n",
    "  layer2_weights = tf.Variable(tf.truncated_normal(\n",
    "      [patch_size, patch_size, depth, depth], stddev=0.1))\n",
    "  layer2_biases = tf.Variable(tf.constant(1.0, shape=[depth]))\n",
    "  layer3_weights = tf.Variable(tf.truncated_normal(\n",
    "      [image_size // 4 * image_size // 4 * depth, num_hidden], stddev=0.1))\n",
    "  layer3_biases = tf.Variable(tf.constant(1.0, shape=[num_hidden]))\n",
    "  layer4_weights = tf.Variable(tf.truncated_normal(\n",
    "      [num_hidden, num_labels], stddev=0.1))\n",
    "  layer4_biases = tf.Variable(tf.constant(1.0, shape=[num_labels]))\n",
    "  \n",
    "  # Model.\n",
    "  def model(data):\n",
    "    conv = tf.nn.conv2d(data, layer1_weights, [1, 2, 2, 1], padding='SAME')\n",
    "    hidden = tf.nn.relu(conv + layer1_biases)\n",
    "    conv = tf.nn.conv2d(hidden, layer2_weights, [1, 2, 2, 1], padding='SAME')\n",
    "    hidden = tf.nn.relu(conv + layer2_biases)\n",
    "    shape = hidden.get_shape().as_list()\n",
    "    reshape = tf.reshape(hidden, [shape[0], shape[1] * shape[2] * shape[3]])\n",
    "    hidden = tf.nn.relu(tf.matmul(reshape, layer3_weights) + layer3_biases)\n",
    "    return tf.matmul(hidden, layer4_weights) + layer4_biases\n",
    "  \n",
    "  # Training computation.\n",
    "  logits = model(tf_train_dataset)\n",
    "  loss = tf.reduce_mean(\n",
    "    tf.nn.softmax_cross_entropy_with_logits(labels=tf_train_labels, logits=logits))\n",
    "    \n",
    "  # Optimizer.\n",
    "  optimizer = tf.train.GradientDescentOptimizer(0.05).minimize(loss)\n",
    "  \n",
    "  # Predictions for the training, validation, and test data.\n",
    "  train_prediction = tf.nn.softmax(logits)\n",
    "  valid_prediction = tf.nn.softmax(model(tf_valid_dataset))\n",
    "  test_prediction = tf.nn.softmax(model(tf_test_dataset))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "cellView": "both",
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "output_extras": [
      {
       "item_id": 37
      }
     ]
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 63292,
     "status": "ok",
     "timestamp": 1446658966251,
     "user": {
      "color": "",
      "displayName": "",
      "isAnonymous": false,
      "isMe": true,
      "permissionId": "",
      "photoUrl": "",
      "sessionId": "0",
      "userId": ""
     },
     "user_tz": 480
    },
    "id": "noKFb2UovVFR",
    "outputId": "28941338-2ef9-4088-8bd1-44295661e628"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Initialized\n",
      "Minibatch loss at step 0: 5.451924\n",
      "Minibatch accuracy: 6.2%\n",
      "Validation accuracy: 11.0%\n",
      "Minibatch loss at step 50: 1.837906\n",
      "Minibatch accuracy: 37.5%\n",
      "Validation accuracy: 45.0%\n",
      "Minibatch loss at step 100: 1.260948\n",
      "Minibatch accuracy: 62.5%\n",
      "Validation accuracy: 63.5%\n",
      "Minibatch loss at step 150: 0.389143\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 74.1%\n",
      "Minibatch loss at step 200: 0.951955\n",
      "Minibatch accuracy: 75.0%\n",
      "Validation accuracy: 76.3%\n",
      "Minibatch loss at step 250: 1.175404\n",
      "Minibatch accuracy: 62.5%\n",
      "Validation accuracy: 76.0%\n",
      "Minibatch loss at step 300: 0.264862\n",
      "Minibatch accuracy: 93.8%\n",
      "Validation accuracy: 78.6%\n",
      "Minibatch loss at step 350: 0.614283\n",
      "Minibatch accuracy: 93.8%\n",
      "Validation accuracy: 75.5%\n",
      "Minibatch loss at step 400: 0.323176\n",
      "Minibatch accuracy: 93.8%\n",
      "Validation accuracy: 79.5%\n",
      "Minibatch loss at step 450: 0.904079\n",
      "Minibatch accuracy: 75.0%\n",
      "Validation accuracy: 78.5%\n",
      "Minibatch loss at step 500: 0.843628\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 80.1%\n",
      "Minibatch loss at step 550: 0.749234\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 79.9%\n",
      "Minibatch loss at step 600: 0.393951\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 81.0%\n",
      "Minibatch loss at step 650: 0.963515\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 81.9%\n",
      "Minibatch loss at step 700: 0.839589\n",
      "Minibatch accuracy: 75.0%\n",
      "Validation accuracy: 81.8%\n",
      "Minibatch loss at step 750: 0.058098\n",
      "Minibatch accuracy: 100.0%\n",
      "Validation accuracy: 82.6%\n",
      "Minibatch loss at step 800: 0.509145\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 82.0%\n",
      "Minibatch loss at step 850: 0.738427\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 81.7%\n",
      "Minibatch loss at step 900: 0.780695\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 82.5%\n",
      "Minibatch loss at step 950: 0.430081\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 82.9%\n",
      "Minibatch loss at step 1000: 0.393164\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 82.4%\n",
      "Minibatch loss at step 1050: 0.584781\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 81.7%\n",
      "Minibatch loss at step 1100: 0.635512\n",
      "Minibatch accuracy: 75.0%\n",
      "Validation accuracy: 83.4%\n",
      "Minibatch loss at step 1150: 0.586390\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 83.7%\n",
      "Minibatch loss at step 1200: 1.127903\n",
      "Minibatch accuracy: 56.2%\n",
      "Validation accuracy: 83.6%\n",
      "Minibatch loss at step 1250: 0.566694\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 83.7%\n",
      "Minibatch loss at step 1300: 0.204646\n",
      "Minibatch accuracy: 93.8%\n",
      "Validation accuracy: 82.8%\n",
      "Minibatch loss at step 1350: 0.855479\n",
      "Minibatch accuracy: 68.8%\n",
      "Validation accuracy: 83.8%\n",
      "Minibatch loss at step 1400: 0.387852\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 84.1%\n",
      "Minibatch loss at step 1450: 0.375003\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 84.4%\n",
      "Minibatch loss at step 1500: 0.719775\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 83.7%\n",
      "Minibatch loss at step 1550: 0.548825\n",
      "Minibatch accuracy: 93.8%\n",
      "Validation accuracy: 83.6%\n",
      "Minibatch loss at step 1600: 0.973149\n",
      "Minibatch accuracy: 75.0%\n",
      "Validation accuracy: 84.1%\n",
      "Minibatch loss at step 1650: 0.646299\n",
      "Minibatch accuracy: 75.0%\n",
      "Validation accuracy: 83.7%\n",
      "Minibatch loss at step 1700: 0.434923\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 84.2%\n",
      "Minibatch loss at step 1750: 0.430219\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 84.4%\n",
      "Minibatch loss at step 1800: 0.257086\n",
      "Minibatch accuracy: 93.8%\n",
      "Validation accuracy: 84.8%\n",
      "Minibatch loss at step 1850: 0.716450\n",
      "Minibatch accuracy: 75.0%\n",
      "Validation accuracy: 84.8%\n",
      "Minibatch loss at step 1900: 0.317823\n",
      "Minibatch accuracy: 93.8%\n",
      "Validation accuracy: 84.6%\n",
      "Minibatch loss at step 1950: 0.564497\n",
      "Minibatch accuracy: 68.8%\n",
      "Validation accuracy: 85.3%\n",
      "Minibatch loss at step 2000: 0.055026\n",
      "Minibatch accuracy: 100.0%\n",
      "Validation accuracy: 84.9%\n",
      "Minibatch loss at step 2050: 0.733915\n",
      "Minibatch accuracy: 68.8%\n",
      "Validation accuracy: 85.0%\n",
      "Minibatch loss at step 2100: 0.277443\n",
      "Minibatch accuracy: 93.8%\n",
      "Validation accuracy: 84.8%\n",
      "Minibatch loss at step 2150: 0.408625\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 85.0%\n",
      "Minibatch loss at step 2200: 0.279579\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 85.2%\n",
      "Minibatch loss at step 2250: 0.504206\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 85.5%\n",
      "Minibatch loss at step 2300: 0.616633\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 85.0%\n",
      "Minibatch loss at step 2350: 0.217243\n",
      "Minibatch accuracy: 93.8%\n",
      "Validation accuracy: 85.4%\n",
      "Minibatch loss at step 2400: 0.588914\n",
      "Minibatch accuracy: 75.0%\n",
      "Validation accuracy: 85.3%\n",
      "Minibatch loss at step 2450: 0.566438\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 85.5%\n",
      "Minibatch loss at step 2500: 0.731982\n",
      "Minibatch accuracy: 75.0%\n",
      "Validation accuracy: 85.3%\n",
      "Minibatch loss at step 2550: 0.557487\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 85.3%\n",
      "Minibatch loss at step 2600: 0.283131\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 85.8%\n",
      "Minibatch loss at step 2650: 0.442248\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 85.7%\n",
      "Minibatch loss at step 2700: 0.362546\n",
      "Minibatch accuracy: 93.8%\n",
      "Validation accuracy: 85.8%\n",
      "Minibatch loss at step 2750: 1.133024\n",
      "Minibatch accuracy: 75.0%\n",
      "Validation accuracy: 85.7%\n",
      "Minibatch loss at step 2800: 0.490875\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 85.8%\n",
      "Minibatch loss at step 2850: 0.079450\n",
      "Minibatch accuracy: 100.0%\n",
      "Validation accuracy: 86.0%\n",
      "Minibatch loss at step 2900: 0.370350\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 85.6%\n",
      "Minibatch loss at step 2950: 0.406407\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 86.4%\n",
      "Minibatch loss at step 3000: 0.856650\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 86.1%\n",
      "Test accuracy: 92.3%\n",
      "Wall time: 45.5 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "num_steps = 3001\n",
    "\n",
    "with tf.Session(graph=graph) as session:\n",
    "  tf.global_variables_initializer().run()\n",
    "  print('Initialized')\n",
    "  for step in range(num_steps):\n",
    "    offset = (step * batch_size) % (train_labels.shape[0] - batch_size)\n",
    "    batch_data = train_dataset[offset:(offset + batch_size), :, :, :]\n",
    "    batch_labels = train_labels[offset:(offset + batch_size), :]\n",
    "    feed_dict = {tf_train_dataset : batch_data, tf_train_labels : batch_labels}\n",
    "    _, l, predictions = session.run(\n",
    "      [optimizer, loss, train_prediction], feed_dict=feed_dict)\n",
    "    if (step % 50 == 0):\n",
    "      print('Minibatch loss at step %d: %f' % (step, l))\n",
    "      print('Minibatch accuracy: %.1f%%' % accuracy(predictions, batch_labels))\n",
    "      print('Validation accuracy: %.1f%%' % accuracy(\n",
    "        valid_prediction.eval(), valid_labels))\n",
    "  print('Test accuracy: %.1f%%' % accuracy(test_prediction.eval(), test_labels))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "KedKkn4EutIK"
   },
   "source": [
    "---\n",
    "Problem 1\n",
    "---------\n",
    "\n",
    "The convolutional model above uses convolutions with stride 2 to reduce the dimensionality. Replace the strides by a max pooling operation (`nn.max_pool()`) of stride 2 and kernel size 2.\n",
    "\n",
    "---"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "cellView": "both",
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "IZYv70SvvOan"
   },
   "outputs": [],
   "source": [
    "batch_size = 16\n",
    "patch_size = 5\n",
    "depth = 16\n",
    "num_hidden = 64\n",
    "\n",
    "graph = tf.Graph()\n",
    "\n",
    "with graph.as_default():\n",
    "\n",
    "  # Input data.\n",
    "  tf_train_dataset = tf.placeholder(\n",
    "    tf.float32, shape=(batch_size, image_size, image_size, num_channels))\n",
    "  tf_train_labels = tf.placeholder(tf.float32, shape=(batch_size, num_labels))\n",
    "  tf_valid_dataset = tf.constant(valid_dataset)\n",
    "  tf_test_dataset = tf.constant(test_dataset)\n",
    "  \n",
    "  # Variables.\n",
    "  layer1_weights = tf.Variable(tf.truncated_normal(\n",
    "      [patch_size, patch_size, num_channels, depth], stddev=0.1))\n",
    "  layer1_biases = tf.Variable(tf.zeros([depth]))\n",
    "  layer2_weights = tf.Variable(tf.truncated_normal(\n",
    "      [patch_size, patch_size, depth, depth], stddev=0.1))\n",
    "  layer2_biases = tf.Variable(tf.constant(1.0, shape=[depth]))\n",
    "  layer3_weights = tf.Variable(tf.truncated_normal(\n",
    "      [image_size // 4 * image_size // 4 * depth, num_hidden], stddev=0.1))\n",
    "  layer3_biases = tf.Variable(tf.constant(1.0, shape=[num_hidden]))\n",
    "  layer4_weights = tf.Variable(tf.truncated_normal(\n",
    "      [num_hidden, num_labels], stddev=0.1))\n",
    "  layer4_biases = tf.Variable(tf.constant(1.0, shape=[num_labels]))\n",
    "  \n",
    "  # Model.\n",
    "  def model(data):\n",
    "    conv = tf.nn.conv2d(data, layer1_weights, [1, 1, 1, 1], padding='SAME')\n",
    "    hidden = tf.nn.relu(conv + layer1_biases)\n",
    "    pool = tf.nn.max_pool(conv, [1, 2, 2, 1], [1, 2, 2, 1], padding='SAME')\n",
    "    conv = tf.nn.conv2d(pool, layer2_weights, [1, 1, 1, 1], padding='SAME')\n",
    "    hidden = tf.nn.relu(conv + layer2_biases)\n",
    "    pool = tf.nn.max_pool(conv, [1, 2, 2, 1], [1, 2, 2, 1], padding='SAME')\n",
    "    shape = pool.get_shape().as_list()\n",
    "    reshape = tf.reshape(pool, [shape[0], shape[1] * shape[2] * shape[3]])\n",
    "    hidden = tf.nn.relu(tf.matmul(reshape, layer3_weights) + layer3_biases)\n",
    "    return tf.matmul(hidden, layer4_weights) + layer4_biases\n",
    "  \n",
    "  # Training computation.\n",
    "  logits = model(tf_train_dataset)\n",
    "  loss = tf.reduce_mean(\n",
    "    tf.nn.softmax_cross_entropy_with_logits(labels=tf_train_labels, logits=logits))\n",
    "    \n",
    "  # Optimizer.\n",
    "  optimizer = tf.train.GradientDescentOptimizer(0.05).minimize(loss)\n",
    "  \n",
    "  # Predictions for the training, validation, and test data.\n",
    "  train_prediction = tf.nn.softmax(logits)\n",
    "  valid_prediction = tf.nn.softmax(model(tf_valid_dataset))\n",
    "  test_prediction = tf.nn.softmax(model(tf_test_dataset))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "cellView": "both",
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "output_extras": [
      {
       "item_id": 37
      }
     ]
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 63292,
     "status": "ok",
     "timestamp": 1446658966251,
     "user": {
      "color": "",
      "displayName": "",
      "isAnonymous": false,
      "isMe": true,
      "permissionId": "",
      "photoUrl": "",
      "sessionId": "0",
      "userId": ""
     },
     "user_tz": 480
    },
    "id": "noKFb2UovVFR",
    "outputId": "28941338-2ef9-4088-8bd1-44295661e628"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Minibatch loss at step 125000: 0.465101\n",
      "Minibatch accuracy: 75.0%\n",
      "Validation accuracy: 89.0%\n",
      "[1 9 3 8 9 3 9 7 0 7 3 4 9 2 4 2 4 2 7 1 1 4 8 0 1 7 8 2 9 8 9 3 4 6 3]\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAD8CAYAAABn919SAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nyBsAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2820d194e80>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAD8CAYAAACb4nSYAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nKQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2820d2806a0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2820e375cc0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test accuracy: 95.0%\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAD8CAYAAACb4nSYAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nvkaj0awh/n9j36/KxXHuwwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2820d1d5f60>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAD8CAYAAACb4nSYAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nKQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2820e37e278>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "num_steps = 125001\n",
    "disp_inc = 500\n",
    "\n",
    "with tf.Session(graph=graph) as session:\n",
    "  tf.global_variables_initializer().run()\n",
    "  print('Initialized')\n",
    "  track_progress = []\n",
    "  track_loss = []\n",
    "  for step in range(num_steps):\n",
    "    offset = (step * batch_size) % (train_labels.shape[0] - batch_size)\n",
    "    batch_data = train_dataset[offset:(offset + batch_size), :, :, :]\n",
    "    batch_labels = train_labels[offset:(offset + batch_size), :]\n",
    "    feed_dict = {tf_train_dataset : batch_data, tf_train_labels : batch_labels}\n",
    "    _, l, predictions = session.run(\n",
    "      [optimizer, loss, train_prediction], feed_dict=feed_dict)\n",
    "    if (step % disp_inc == 0):\n",
    "      val_pred = valid_prediction.eval()\n",
    "      valid_acc = accuracy(val_pred, valid_labels)\n",
    "      track_progress.append(valid_acc)\n",
    "      track_loss.append(l)\n",
    "        \n",
    "      clear_output(wait=True)\n",
    "    \n",
    "      print('Minibatch loss at step %d: %f' % (step, l))\n",
    "      print('Minibatch accuracy: %.1f%%' % accuracy(predictions, batch_labels))\n",
    "      print('Validation accuracy: %.1f%%' % valid_acc)\n",
    "      print(np.argmax(val_pred[0:35], 1))\n",
    "      \n",
    "      plt.plot([disp_inc * i for i in range(len(track_progress))], track_progress)\n",
    "      plt.show()\n",
    "      drop_some = track_progress[(len(track_progress) + 5) // 10:]\n",
    "      plt.plot([disp_inc * i for i in range(len(track_progress) - len(drop_some), len(track_progress))], drop_some)\n",
    "      plt.show()\n",
    "      plt.plot([disp_inc * i for i in range(len(track_loss))], track_loss)\n",
    "      plt.show()\n",
    "\n",
    "  print('Test accuracy: %.1f%%' % accuracy(test_prediction.eval(), test_labels))\n",
    "  last_bit = track_progress[-30:]\n",
    "  plt.plot([disp_inc * i for i in range(len(track_progress) - len(last_bit), len(track_progress))], last_bit)\n",
    "  plt.show()\n",
    "  last_bit = track_progress[(len(track_progress) + 5) // 10:]\n",
    "  plt.plot([disp_inc * i for i in range(len(track_progress) - len(last_bit), len(track_progress))], last_bit)\n",
    "  plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "klf21gpbAgb-"
   },
   "source": [
    "---\n",
    "Problem 2\n",
    "---------\n",
    "\n",
    "Try to get the best performance you can using a convolutional net. Look for example at the classic [LeNet5](http://yann.lecun.com/exdb/lenet/) architecture, adding Dropout, and/or adding learning rate decay.\n",
    "\n",
    "---"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 225,
   "metadata": {
    "cellView": "both",
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "IZYv70SvvOan"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.4714045207910317\n",
      "0.5\n",
      "0.16666666666666666\n",
      "0.16666666666666666\n",
      "0.16666666666666666\n",
      "0.3535533905932738\n",
      "0.11785113019775792\n",
      "0.11785113019775792\n",
      "0.11785113019775792\n",
      "0.25\n",
      "0.08333333333333333\n",
      "0.08333333333333333\n",
      "0.03571428571428571\n",
      "0.09128709291752768\n",
      "0.18257418583505536\n"
     ]
    }
   ],
   "source": [
    "batch_size = 16\n",
    "patch_size = 5\n",
    "depths = [8, 16, 32]\n",
    "num_hidden = [240, 60]\n",
    "loss_coeff = 0.03\n",
    "keep_prob = 0.6\n",
    "\n",
    "graph = tf.Graph()\n",
    "\n",
    "with graph.as_default():\n",
    "\n",
    "  # Input data.\n",
    "  tf_train_dataset = tf.placeholder(\n",
    "    tf.float32, shape=(batch_size, image_size, image_size, num_channels))\n",
    "  tf_train_labels = tf.placeholder(tf.float32, shape=(batch_size, num_labels))\n",
    "  tf_valid_dataset = tf.constant(valid_dataset)\n",
    "  tf_test_dataset = tf.constant(test_dataset)\n",
    "\n",
    "  def activate(activations, training):\n",
    "    res = tf.nn.relu(activations)\n",
    "    if training:\n",
    "      res = tf.nn.dropout(res, keep_prob)\n",
    "    return res\n",
    "\n",
    "  def xavier(shape):\n",
    "    stddev =  math.sqrt(2.0 / np.prod(shape[:len(shape) - 1]))\n",
    "    print(stddev)\n",
    "    return stddev\n",
    "\n",
    "  def init_weights(shape):\n",
    "    return tf.Variable(tf.truncated_normal(shape, stddev=xavier(shape)))\n",
    "\n",
    "  def conv_var(patch_size, in_depth, out_depth):\n",
    "    return init_weights([patch_size, patch_size, in_depth, out_depth])\n",
    "\n",
    "  def make_inception(in_depth, out_depth):\n",
    "    one_weights = [ conv_var(1, in_depth, out_depth) for _ in range(4) ]\n",
    "    one_biases = [ tf.Variable(tf.zeros([out_depth])) for _ in range(4) ]\n",
    "    three_weights = conv_var(3, out_depth, out_depth)\n",
    "    three_biases = tf.Variable(tf.zeros([out_depth]))\n",
    "    five_weights = conv_var(5, out_depth, out_depth)\n",
    "    five_biases = tf.Variable(tf.zeros([out_depth]))\n",
    "    \n",
    "    def app(data):\n",
    "      res_1 = activate(tf.nn.conv2d(data, one_weights[0], [1, 1, 1, 1], padding='SAME') + one_biases[0])\n",
    "      res_2 = activate(tf.nn.conv2d(data, one_weights[1], [1, 1, 1, 1], padding='SAME') + one_biases[1])\n",
    "      res_2 = activate(tf.nn.conv2d(res_2, three_weights, [1, 1, 1, 1], padding='SAME') + three_biases)\n",
    "      res_3 = activate(tf.nn.conv2d(data, one_weights[2], [1, 1, 1, 1], padding='SAME') + one_biases[2])\n",
    "      res_3 = activate(tf.nn.conv2d(res_3, five_weights, [1, 1, 1, 1], padding='SAME') + five_biases)\n",
    "      res_4 = tf.nn.avg_pool(data, [1, 3, 3, 1], [1, 1, 1, 1], padding='SAME')\n",
    "      res_4 = activate(tf.nn.conv2d(res_4, one_weights[3], [1, 1, 1, 1,], padding='SAME') + one_biases[3])\n",
    "    \n",
    "      return tf.concat([res_1, res_2, res_3, res_4], 3)\n",
    "    \n",
    "    return app\n",
    "\n",
    "  def make_deep_conv(in_depth, out_depth):  \n",
    "    layer_ws = [ conv_var(patch_size, in_depth, out_depth) ]\n",
    "    layer_ws = layer_ws + [ conv_var(p, out_depth, out_depth) for p in [1, 3, 3] ]\n",
    "    layer_bs = [ tf.Variable(tf.zeros([out_depth])) for _ in range(4) ]\n",
    "    \n",
    "    def app(data, training):\n",
    "      res = data\n",
    "      for w, b in zip(layer_ws, layer_bs):\n",
    "        res = activate(tf.nn.conv2d(res, w, [1, 1, 1, 1], padding='SAME') + b, training)\n",
    "      return res\n",
    "    \n",
    "    return app, layer_ws\n",
    "  \n",
    "  # Variables.\n",
    "  padded_depths = [num_channels] + depths\n",
    "  weights = []\n",
    "  deep_convs = []\n",
    "  for app, ws in [ make_deep_conv(padded_depths[i], padded_depths[i + 1]) for i in range(len(depths)) ]:\n",
    "    weights = weights + ws\n",
    "    deep_convs.append(app) \n",
    "  layer3_weights = init_weights([image_size // 4 * image_size // 4 * depths[-1], num_hidden[0]])\n",
    "  layer3_biases = tf.Variable(tf.constant(1.0, shape=[num_hidden[0]]))  \n",
    "  layer4_weights = init_weights([num_hidden[0], num_hidden[1]])\n",
    "  layer4_biases = tf.Variable(tf.constant(1.0, shape=[num_hidden[1]]))\n",
    "  layer5_weights = init_weights([num_hidden[1], num_labels])\n",
    "  layer5_biases = tf.Variable(tf.constant(1.0, shape=[num_labels]))\n",
    "    \n",
    "  weights = weights + [layer3_weights, layer4_weights, layer4_weights]\n",
    "  \n",
    "  # Model.\n",
    "  def model(data, training=False):\n",
    "    pool = data\n",
    "    for deep_conv in deep_convs:\n",
    "      res = deep_conv(pool, training)\n",
    "      if res.get_shape().as_list()[1] % 2 == 0:\n",
    "        pool = tf.nn.max_pool(res, [1, 2, 2, 1], [1, 2, 2, 1], padding='SAME')\n",
    "      else:\n",
    "        pool = res\n",
    "    shape = pool.get_shape().as_list()\n",
    "    reshape = tf.reshape(pool, [shape[0], shape[1] * shape[2] * shape[3]])\n",
    "    hidden = activate(tf.matmul(reshape, layer3_weights) + layer3_biases, training)\n",
    "    hidden = activate(tf.matmul(hidden, layer4_weights) + layer4_biases, training)\n",
    "    return tf.matmul(hidden, layer5_weights) + layer5_biases\n",
    "  \n",
    "  # Training computation.\n",
    "  logits = model(tf_train_dataset, training=False)\n",
    "  loss = (tf.reduce_mean(\n",
    "          tf.nn.softmax_cross_entropy_with_logits(labels=tf_train_labels, logits=logits)))\n",
    "  for w in weights:\n",
    "    loss = loss + tf.nn.l2_loss(w) * loss_coeff\n",
    "    \n",
    "  # Optimizer.\n",
    "  global_step = tf.Variable(0)\n",
    "  learning_rate = tf.train.exponential_decay(0.03, global_step, 1000, 0.95)\n",
    "  #learning_rate = tf.train.polynomial_decay(0.05, global_step, num_steps)\n",
    "  optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss, global_step=global_step)\n",
    "  \n",
    "  # Predictions for the training, validation, and test data.\n",
    "  train_prediction = tf.nn.softmax(logits)\n",
    "  valid_prediction = tf.nn.softmax(model(tf_valid_dataset))\n",
    "  test_prediction = tf.nn.softmax(model(tf_test_dataset))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 238,
   "metadata": {
    "cellView": "both",
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "IZYv70SvvOan"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.4142135623730951\n",
      "1.4142135623730951\n",
      "1.4142135623730951\n",
      "1.4142135623730951\n",
      "0.23570226039551584\n",
      "0.1414213562373095\n",
      "0.3535533905932738\n",
      "0.3535533905932738\n",
      "0.3535533905932738\n",
      "0.3535533905932738\n",
      "0.16666666666666666\n",
      "0.1\n",
      "0.25\n",
      "0.25\n",
      "0.25\n",
      "0.25\n",
      "0.11785113019775792\n",
      "0.07071067811865475\n",
      "0.025253813613805267\n",
      "0.14433756729740643\n",
      "0.25\n"
     ]
    }
   ],
   "source": [
    "batch_size = 16\n",
    "patch_size = 5\n",
    "depths = [4, 8, 16]\n",
    "num_hidden = [96, 32]\n",
    "loss_coeff = 0.03\n",
    "\n",
    "graph = tf.Graph()\n",
    "\n",
    "with graph.as_default():\n",
    "\n",
    "  # Input data.\n",
    "  tf_train_dataset = tf.placeholder(\n",
    "    tf.float32, shape=(batch_size, image_size, image_size, num_channels))\n",
    "  tf_train_labels = tf.placeholder(tf.float32, shape=(batch_size, num_labels))\n",
    "  tf_valid_dataset = tf.constant(valid_dataset)\n",
    "  tf_test_dataset = tf.constant(test_dataset)\n",
    "\n",
    "  def activate(activations):\n",
    "    return tf.nn.relu(activations)\n",
    "\n",
    "  def xavier(shape):\n",
    "    stddev = math.sqrt(2 / np.prod(shape[:len(shape) - 1]))\n",
    "    print(stddev)\n",
    "    return stddev\n",
    "\n",
    "  def init_weights(shape):\n",
    "    return tf.Variable(tf.truncated_normal(shape, stddev=xavier(shape)))\n",
    "\n",
    "  def conv_var(patch_size, in_depth, out_depth):\n",
    "    return init_weights([patch_size, patch_size, in_depth, out_depth])\n",
    "\n",
    "  def make_inception(in_depth, out_depth):\n",
    "    one_weights = [ conv_var(1, in_depth, out_depth) for _ in range(4) ]\n",
    "    one_biases = [ tf.Variable(tf.zeros([out_depth])) for _ in range(4) ]\n",
    "    three_weights = conv_var(3, out_depth, out_depth)\n",
    "    three_biases = tf.Variable(tf.zeros([out_depth]))\n",
    "    five_weights = conv_var(5, out_depth, out_depth)\n",
    "    five_biases = tf.Variable(tf.zeros([out_depth]))\n",
    "    \n",
    "    def app(data):\n",
    "      res_1 = activate(tf.nn.conv2d(data, one_weights[0], [1, 1, 1, 1], padding='SAME') + one_biases[0])\n",
    "      res_2 = activate(tf.nn.conv2d(data, one_weights[1], [1, 1, 1, 1], padding='SAME') + one_biases[1])\n",
    "      res_2 = activate(tf.nn.conv2d(res_2, three_weights, [1, 1, 1, 1], padding='SAME') + three_biases)\n",
    "      res_3 = activate(tf.nn.conv2d(data, one_weights[2], [1, 1, 1, 1], padding='SAME') + one_biases[2])\n",
    "      res_3 = activate(tf.nn.conv2d(res_3, five_weights, [1, 1, 1, 1], padding='SAME') + five_biases)\n",
    "      res_4 = tf.nn.avg_pool(data, [1, 3, 3, 1], [1, 1, 1, 1], padding='SAME')\n",
    "      res_4 = activate(tf.nn.conv2d(res_4, one_weights[3], [1, 1, 1, 1,], padding='SAME') + one_biases[3])\n",
    "    \n",
    "      return tf.concat([res_1, res_2, res_3, res_4], 3)\n",
    "    \n",
    "    return app, one_weights + [three_weights, five_weights]\n",
    "  \n",
    "  # Variables.\n",
    "  weights = []\n",
    "  incep_1, ws = make_inception(num_channels, depths[0])\n",
    "  weights = weights + ws\n",
    "  incep_2, ws = make_inception(4 * depths[0], depths[1])\n",
    "  weights = weights + ws\n",
    "  incep_3, ws = make_inception(4 * depths[1], depths[2])\n",
    "  weights = weights + ws\n",
    "  layer3_weights = init_weights([image_size // 4 * image_size // 4 * depths[2] * 4, num_hidden[0]])\n",
    "  layer3_biases = tf.Variable(tf.constant(1.0, shape=[num_hidden[0]]))  \n",
    "  layer4_weights = init_weights([num_hidden[0], num_hidden[1]])\n",
    "  layer4_biases = tf.Variable(tf.constant(1.0, shape=[num_hidden[1]]))\n",
    "  layer5_weights = init_weights([num_hidden[1], num_labels])\n",
    "  layer5_biases = tf.Variable(tf.constant(1.0, shape=[num_labels]))\n",
    "  weights = weights + [layer3_weights, layer4_weights, layer5_weights]\n",
    "  \n",
    "  # Model.\n",
    "  def model(data):\n",
    "    res = incep_1(data)\n",
    "    res = incep_2(res)\n",
    "    pool = tf.nn.max_pool(res, [1, 2, 2, 1], [1, 2, 2, 1], padding='SAME')\n",
    "    res = incep_3(pool)\n",
    "    pool = tf.nn.max_pool(res, [1, 2, 2, 1], [1, 2, 2, 1], padding='SAME')\n",
    "    shape = pool.get_shape().as_list()\n",
    "    reshape = tf.reshape(pool, [shape[0], shape[1] * shape[2] * shape[3]])\n",
    "    hidden = tf.nn.relu(tf.matmul(reshape, layer3_weights) + layer3_biases)\n",
    "    hidden = tf.nn.relu(tf.matmul(hidden, layer4_weights) + layer4_biases)\n",
    "    return tf.matmul(hidden, layer5_weights) + layer5_biases\n",
    "  \n",
    "  # Training computation.\n",
    "  logits = model(tf_train_dataset)\n",
    "  loss = tf.reduce_mean(\n",
    "    tf.nn.softmax_cross_entropy_with_logits(labels=tf_train_labels, logits=logits))\n",
    "  for w in weights:\n",
    "    loss = loss + loss_coeff * tf.nn.l2_loss(w)\n",
    "    \n",
    "  # Optimizer.\n",
    "  global_step = tf.Variable(0)\n",
    "  learning_rate = tf.train.exponential_decay(0.1, global_step, 500, 0.95)\n",
    "  #learning_rate = tf.train.polynomial_decay(0.05, global_step, num_steps)\n",
    "  optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss, global_step=global_step)\n",
    "  \n",
    "  # Predictions for the training, validation, and test data.\n",
    "  train_prediction = tf.nn.softmax(logits)\n",
    "  valid_prediction = tf.nn.softmax(model(tf_valid_dataset))\n",
    "  test_prediction = tf.nn.softmax(model(tf_test_dataset))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.282842712474619\n",
      "0.07071067811865475\n",
      "0.05\n",
      "0.025253813613805267\n",
      "0.044721359549995794\n",
      "0.08164965809277261\n",
      "0.18257418583505536\n"
     ]
    }
   ],
   "source": [
    "batch_size = 16\n",
    "patches = [5, 5, 5]\n",
    "depths = [16, 32, 64]\n",
    "hiddens = [1000, 300, 60]\n",
    "\n",
    "graph = tf.Graph()\n",
    "\n",
    "with graph.as_default(), tf.device('/cpu:0'):\n",
    "\n",
    "  # Input data.\n",
    "  tf_train_dataset = tf.placeholder(\n",
    "    tf.float32, shape=(batch_size, image_size, image_size, num_channels))\n",
    "  tf_train_labels = tf.placeholder(tf.float32, shape=(batch_size, num_labels))\n",
    "  tf_valid_dataset = tf.constant(valid_dataset)\n",
    "  tf_test_dataset = tf.constant(test_dataset)\n",
    "  \n",
    "  def activate(activations):\n",
    "    return tf.nn.relu(activations)\n",
    "\n",
    "  def xavier(shape):\n",
    "    stddev = math.sqrt(2 / np.prod(shape[:len(shape) - 1]))\n",
    "    print(stddev)\n",
    "    return stddev\n",
    "\n",
    "  def init_weights(shape):\n",
    "    return tf.Variable(tf.truncated_normal(shape, stddev=xavier(shape)))\n",
    "\n",
    "  def conv_var(patch_size, in_depth, out_depth):\n",
    "    return init_weights([patch_size, patch_size, in_depth, out_depth])\n",
    "    \n",
    "  # Variables.\n",
    "  padded = [num_channels] + depths\n",
    "  conv_ws = [ conv_var(patches[i], padded[i], padded[i + 1]) for i in range(len(depths)) ]\n",
    "  conv_bs = [ tf.Variable(tf.zeros([depth])) for depth in depths ]\n",
    "  padded = [image_size // 4 * image_size // 4 * depths[-1]] + hiddens + [num_labels]\n",
    "  hidden_ws = [ init_weights([padded[i], padded[i+1]]) for i in range(len(hiddens) + 1) ]\n",
    "  hidden_bs = [ tf.Variable(tf.zeros([size])) for size in hiddens + [num_labels]]\n",
    "  \n",
    "  # Model.\n",
    "  def model(data):\n",
    "    pool = data\n",
    "    for i in range(len(depths)):\n",
    "      conv = activate(tf.nn.conv2d(pool, conv_ws[i], [1, 1, 1, 1], padding='SAME') + conv_bs[i])\n",
    "      if len(depths) - i < 3:\n",
    "        pool = tf.nn.max_pool(conv, [1, 2, 2, 1], [1, 2, 2, 1], padding='SAME')\n",
    "      else:\n",
    "        pool = conv\n",
    "    shape = pool.get_shape().as_list()\n",
    "    reshape = tf.reshape(pool, [shape[0], shape[1] * shape[2] * shape[3]])\n",
    "    res = tf.matmul(reshape, hidden_ws[0]) + hidden_bs[0]\n",
    "    for i in range(1, len(hidden_ws)):\n",
    "      act = activate(res)\n",
    "      res = tf.matmul(act, hidden_ws[i]) + hidden_bs[i]\n",
    "    return res\n",
    "  \n",
    "  # Training computation.\n",
    "  logits = model(tf_train_dataset)\n",
    "  loss = tf.reduce_mean(\n",
    "    tf.nn.softmax_cross_entropy_with_logits(labels=tf_train_labels, logits=logits))\n",
    "    \n",
    "    # Optimizer.\n",
    "  global_step = tf.Variable(0)\n",
    "  learning_rate = tf.train.exponential_decay(0.1, global_step, 2000, 0.95)\n",
    "  #learning_rate = tf.train.polynomial_decay(0.05, global_step, num_steps)\n",
    "  optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss, global_step=global_step)\n",
    "  \n",
    "  # Predictions for the training, validation, and test data.\n",
    "  train_prediction = tf.nn.softmax(logits)\n",
    "  valid_prediction = tf.nn.softmax(model(tf_valid_dataset))\n",
    "  test_prediction = tf.nn.softmax(model(tf_test_dataset))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "cellView": "both",
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "output_extras": [
      {
       "item_id": 37
      }
     ]
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 63292,
     "status": "ok",
     "timestamp": 1446658966251,
     "user": {
      "color": "",
      "displayName": "",
      "isAnonymous": false,
      "isMe": true,
      "permissionId": "",
      "photoUrl": "",
      "sessionId": "0",
      "userId": ""
     },
     "user_tz": 480
    },
    "id": "noKFb2UovVFR",
    "outputId": "28941338-2ef9-4088-8bd1-44295661e628",
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Minibatch loss at step 4500: 0.442883\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 89.3%\n",
      "[1 9 3 8 9 3 9 7 0 0 3 4 9 2 4 2 4 2 7 1 1 4 8 0 3 7 8 2 9 8 9 9 4 6 3]\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAD8CAYAAABn919SAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nA83pIRyuMAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1fc9f965e80>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAD8CAYAAABn919SAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1fca201e390>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ni4hEIYW7iEgU+v/WOoc4MlbNDAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1fc9ffb7dd8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "num_steps = 5001 #125001\n",
    "disp_inc = 500\n",
    "\n",
    "with tf.Session(graph=graph) as session:\n",
    "  tf.global_variables_initializer().run()\n",
    "  print('Initialized')\n",
    "  track_progress = []\n",
    "  track_loss = []\n",
    "  for step in range(num_steps):\n",
    "    try:\n",
    "      offset = (step * batch_size) % (train_labels.shape[0] - batch_size)\n",
    "      batch_data = train_dataset[offset:(offset + batch_size), :, :, :]\n",
    "      batch_labels = train_labels[offset:(offset + batch_size), :]\n",
    "      feed_dict = {tf_train_dataset : batch_data, tf_train_labels : batch_labels}\n",
    "      _, l, predictions = session.run(\n",
    "        [optimizer, loss, train_prediction], feed_dict=feed_dict)\n",
    "      if (step % disp_inc == 0):\n",
    "        val_pred = valid_prediction.eval()\n",
    "        valid_acc = accuracy(val_pred, valid_labels)\n",
    "        track_progress.append(valid_acc)\n",
    "        track_loss.append(l)\n",
    "          \n",
    "        clear_output(wait=True)\n",
    "      \n",
    "        print('Minibatch loss at step %d: %f' % (step, l))\n",
    "        print('Minibatch accuracy: %.1f%%' % accuracy(predictions, batch_labels))\n",
    "        print('Validation accuracy: %.1f%%' % valid_acc)\n",
    "        print(np.argmax(val_pred[0:35], 1))\n",
    "      \n",
    "        plt.plot([disp_inc * i for i in range(len(track_progress))], track_progress)\n",
    "        plt.show()\n",
    "        drop_some = track_progress[(len(track_progress) + 5) // 10:]\n",
    "        plt.plot([disp_inc * i for i in range(len(track_progress) - len(drop_some), len(track_progress))], drop_some)\n",
    "        plt.show()\n",
    "        plt.plot([disp_inc * i for i in range(len(track_loss))], track_loss)\n",
    "        plt.show()\n",
    "        \n",
    "    except KeyboardInterrupt:\n",
    "      print('Stopped early because KeyboardInterrupt')\n",
    "      break\n",
    "\n",
    "  print('Test accuracy: %.1f%%' % accuracy(test_prediction.eval(), test_labels))\n",
    "  last_bit = track_progress[-30:]\n",
    "  plt.plot([disp_inc * i for i in range(len(track_progress) - len(last_bit), len(track_progress))], last_bit)\n",
    "  plt.show()\n",
    "  last_bit = track_progress[(len(track_progress) + 5) // 10:]\n",
    "  plt.plot([disp_inc * i for i in range(len(track_progress) - len(last_bit), len(track_progress))], last_bit)\n",
    "  plt.show()"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "default_view": {},
   "name": "4_convolutions.ipynb",
   "provenance": [],
   "version": "0.3.2",
   "views": {}
  },
  "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}