sec1-4-building-better-cnns-with-regularization.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Better CNNs using regularization\n",
    "\n",
    "Dataset of 60,000 28x28 grayscale images of 10 fashion categories, along with a test set of 10,000 images. This dataset can be used as a drop-in replacement for MNIST. The class labels are:\n",
    "\n",
    "Label Description\n",
    "\n",
    "- 0 T-shirt/top\n",
    "- 1 Trouser\n",
    "- 2 Pullover\n",
    "- 3 Dress\n",
    "- 4 Coat\n",
    "- 5 Sandal\n",
    "- 6 Shirt\n",
    "- 7 Sneaker\n",
    "- 8 Bag\n",
    "- 9 Ankle boot\n",
    "\n",
    "This dataset is available from Keras"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Importing the necessary libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\direc\\Anaconda3\\envs\\recipes-ai\\lib\\site-packages\\h5py\\__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Loading the dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.datasets import fashion_mnist\n",
    "(X_train, y_train), (X_test, y_test) = fashion_mnist.load_data()\n",
    "\n",
    "n_train = y_train.shape[0]\n",
    "n_test = y_test.shape[0]\n",
    "n_categories = 10"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Take a look at the dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2876f1092b0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "categories = {0:'T-shirt/top', 1:'Trouser', 2:'Pullover', 3:'Dress', 4:'Coat',\n",
    "                 5:'Sandal', 6:'Shirt', 7:'Sneaker', 8:'Bag', 9:'Ankle boot'}\n",
    "\n",
    "fig, axes = plt.subplots(nrows=4, ncols=8, figsize=(13,6))\n",
    "for i, ax in enumerate(axes.flatten()):\n",
    "    ax.imshow(X_train[i], cmap='Greys')\n",
    "    ax.set_title(categories[y_train[i]])\n",
    "    ax.axis('off')\n",
    "\n",
    "fig.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Preprocesing the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "## Reshaping and rescaling images:\n",
    "X_train = X_train.reshape(n_train, 28, 28, 1)/255.\n",
    "X_test = X_test.reshape(n_test, 28, 28, 1)/255.\n",
    "\n",
    "## Getting validation and test sets\n",
    "X_val = X_test[:1000]\n",
    "X_test = X_test[1000:]\n",
    "\n",
    "y_val = y_test[:1000]\n",
    "y_test = y_test[1000:]\n",
    "## Here we can use 1000 for test and validation and the reset for training"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Build the input pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "BATCH_SIZE = 128\n",
    "\n",
    "train_dataset = tf.data.Dataset.from_tensor_slices((X_train, y_train))\n",
    "train_dataset = train_dataset.shuffle(buffer_size=10000)\n",
    "train_dataset = train_dataset.batch(BATCH_SIZE)\n",
    "iterator = train_dataset.make_initializable_iterator()\n",
    "next_element = iterator.get_next()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Building the CNN model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "training = tf.placeholder_with_default(False, shape=[])\n",
    "dropout_rate = 0.4\n",
    "def cnn(inputs):\n",
    "    ## 2 - Convolutions + 1 - maxpooling\n",
    "    conv1 = tf.layers.conv2d(inputs, filters=16, kernel_size=3, strides=1,\n",
    "                             padding='same', activation=tf.nn.relu)\n",
    "    conv2 = tf.layers.conv2d(conv1, filters=32, kernel_size=3, strides=1,\n",
    "                             padding='same', activation=tf.nn.relu)\n",
    "    max_pooling1 = tf.layers.max_pooling2d(conv2, pool_size=2, strides=2, padding='valid')\n",
    "    ## 2 - Convolutions + 1 - maxpooling    \n",
    "    conv3 = tf.layers.conv2d(max_pooling1, filters=32, kernel_size=3, strides=1,\n",
    "                             padding='same', activation=tf.nn.relu)\n",
    "    conv4 = tf.layers.conv2d(conv3, filters=64, kernel_size=3, strides=1,\n",
    "                             padding='same', activation=tf.nn.relu)\n",
    "    max_pooling2 = tf.layers.max_pooling2d(conv4, pool_size=2, strides=2, padding='valid')\n",
    "    \n",
    "    dropout1 = tf.layers.dropout(max_pooling2, rate=dropout_rate, training=training)\n",
    "    \n",
    "    ## Flatten + dense + dropout + dense + dense\n",
    "    flatten = tf.layers.flatten(dropout1)\n",
    "    dense1 = tf.layers.dense(flatten, units=64, activation=tf.nn.relu)\n",
    "    dropout2 = tf.layers.dropout(dense1, rate=dropout_rate, training=training)\n",
    "    dense2 = tf.layers.dense(dropout2, units=32, activation=tf.nn.relu)\n",
    "    \n",
    "    ## Output\n",
    "    logits = tf.layers.dense(dense2, units=n_categories)\n",
    "    return logits"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create placeholders to pass values for training and evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X_placeholder_shape = [None] + list(X_train.shape[1:])\n",
    "X = tf.placeholder(tf.float32, shape=X_placeholder_shape)\n",
    "y = tf.placeholder(tf.int32)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Define the loss, optimizer and training operation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "logits = cnn(X)\n",
    "loss = tf.losses.sparse_softmax_cross_entropy(labels=y, logits=logits)\n",
    "optimizer = tf.train.AdamOptimizer() ## here we are changing optimizer\n",
    "training_op = optimizer.minimize(loss)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Getting predictions and metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y_pred_proba = tf.nn.softmax(logits)\n",
    "y_pred = tf.argmax(y_pred_proba, axis=1, output_type=tf.int32)\n",
    "\n",
    "correct = tf.equal(y_pred, y)\n",
    "accuracy = 100*tf.reduce_mean(tf.cast(correct, tf.float32))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Write a function for running the training operation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def train_model(epoch):\n",
    "    iterator.initializer.run()\n",
    "    while True:\n",
    "        try:\n",
    "            X_values, y_values = sess.run(next_element)\n",
    "            sess.run(training_op, feed_dict={X: X_values, y:y_values, training:True})\n",
    "        except tf.errors.OutOfRangeError:\n",
    "            break\n",
    "    train_acc = sess.run(accuracy, feed_dict={X: X_values, y:y_values})\n",
    "    val_acc = sess.run(accuracy, feed_dict={X: X_val, y:y_val})\n",
    "    print('Epoch: {:d}, Training Accuracy: {:0.1f}% | Validation Accuracy {:0.1f}%'.format(epoch, train_acc, val_acc))\n",
    "    return train_acc, val_acc"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Run the computational graph"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 1, Training Accuracy: 90.6% | Validation Accuracy 85.9%\n",
      "Epoch: 2, Training Accuracy: 89.6% | Validation Accuracy 87.8%\n",
      "Epoch: 3, Training Accuracy: 90.6% | Validation Accuracy 89.6%\n",
      "Epoch: 4, Training Accuracy: 91.7% | Validation Accuracy 89.6%\n",
      "Epoch: 5, Training Accuracy: 92.7% | Validation Accuracy 91.0%\n",
      "Epoch: 6, Training Accuracy: 92.7% | Validation Accuracy 91.3%\n",
      "Epoch: 7, Training Accuracy: 93.8% | Validation Accuracy 92.0%\n",
      "Epoch: 8, Training Accuracy: 89.6% | Validation Accuracy 92.5%\n",
      "Epoch: 9, Training Accuracy: 93.8% | Validation Accuracy 92.5%\n",
      "Epoch: 10, Training Accuracy: 96.9% | Validation Accuracy 92.4%\n",
      "Epoch: 11, Training Accuracy: 95.8% | Validation Accuracy 93.1%\n",
      "Epoch: 12, Training Accuracy: 93.8% | Validation Accuracy 93.4%\n",
      "Epoch: 13, Training Accuracy: 94.8% | Validation Accuracy 93.0%\n",
      "Epoch: 14, Training Accuracy: 96.9% | Validation Accuracy 92.8%\n",
      "Epoch: 15, Training Accuracy: 95.8% | Validation Accuracy 93.6%\n",
      "Epoch: 16, Training Accuracy: 95.8% | Validation Accuracy 93.6%\n",
      "Epoch: 17, Training Accuracy: 94.8% | Validation Accuracy 94.1%\n",
      "Epoch: 18, Training Accuracy: 91.7% | Validation Accuracy 93.5%\n",
      "Epoch: 19, Training Accuracy: 96.9% | Validation Accuracy 93.5%\n",
      "Epoch: 20, Training Accuracy: 96.9% | Validation Accuracy 92.9%\n",
      "Epoch: 21, Training Accuracy: 95.8% | Validation Accuracy 93.4%\n",
      "Epoch: 22, Training Accuracy: 96.9% | Validation Accuracy 94.1%\n",
      "Epoch: 23, Training Accuracy: 97.9% | Validation Accuracy 93.2%\n",
      "Epoch: 24, Training Accuracy: 99.0% | Validation Accuracy 93.2%\n",
      "Epoch: 25, Training Accuracy: 99.0% | Validation Accuracy 93.8%\n",
      "Epoch: 26, Training Accuracy: 96.9% | Validation Accuracy 94.0%\n",
      "Epoch: 27, Training Accuracy: 99.0% | Validation Accuracy 93.6%\n",
      "Epoch: 28, Training Accuracy: 95.8% | Validation Accuracy 94.3%\n",
      "Epoch: 29, Training Accuracy: 95.8% | Validation Accuracy 93.5%\n",
      "Epoch: 30, Training Accuracy: 93.8% | Validation Accuracy 94.2%\n",
      "Epoch: 31, Training Accuracy: 100.0% | Validation Accuracy 94.0%\n",
      "Epoch: 32, Training Accuracy: 95.8% | Validation Accuracy 93.8%\n",
      "Epoch: 33, Training Accuracy: 97.9% | Validation Accuracy 94.0%\n",
      "Epoch: 34, Training Accuracy: 94.8% | Validation Accuracy 93.7%\n",
      "Epoch: 35, Training Accuracy: 96.9% | Validation Accuracy 93.4%\n",
      "Epoch: 36, Training Accuracy: 99.0% | Validation Accuracy 93.6%\n",
      "Epoch: 37, Training Accuracy: 99.0% | Validation Accuracy 93.5%\n",
      "Epoch: 38, Training Accuracy: 93.8% | Validation Accuracy 93.7%\n",
      "Epoch: 39, Training Accuracy: 99.0% | Validation Accuracy 93.7%\n",
      "Epoch: 40, Training Accuracy: 96.9% | Validation Accuracy 94.3%\n",
      "Epoch: 41, Training Accuracy: 96.9% | Validation Accuracy 93.9%\n",
      "Epoch: 42, Training Accuracy: 97.9% | Validation Accuracy 94.0%\n",
      "Epoch: 43, Training Accuracy: 95.8% | Validation Accuracy 94.4%\n",
      "Epoch: 44, Training Accuracy: 97.9% | Validation Accuracy 94.1%\n",
      "Epoch: 45, Training Accuracy: 99.0% | Validation Accuracy 94.1%\n",
      "Epoch: 46, Training Accuracy: 99.0% | Validation Accuracy 94.0%\n",
      "Epoch: 47, Training Accuracy: 96.9% | Validation Accuracy 94.6%\n",
      "Epoch: 48, Training Accuracy: 97.9% | Validation Accuracy 94.3%\n",
      "Epoch: 49, Training Accuracy: 99.0% | Validation Accuracy 93.5%\n",
      "Epoch: 50, Training Accuracy: 99.0% | Validation Accuracy 93.9%\n",
      "Done training\n"
     ]
    }
   ],
   "source": [
    "N_EPOCHS = 50\n",
    "accuracy_metrics = np.zeros((N_EPOCHS, 2))\n",
    "with tf.Session() as sess:\n",
    "    tf.global_variables_initializer().run()\n",
    "    for epoch in range(N_EPOCHS):\n",
    "        accuracy_metrics[epoch] = train_model(epoch+1)\n",
    "    print(\"Done training\")\n",
    "    test_preds = sess.run(y_pred_proba, feed_dict={X: X_test[:1500], y:y_test[:1500]})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluation metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2871b4bdb38>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(8,5))\n",
    "ax.plot(np.arange(N_EPOCHS), accuracy_metrics[:,0], 'o-', label='Training')\n",
    "ax.plot(np.arange(N_EPOCHS), accuracy_metrics[:,1], 'o-', label='Validation')\n",
    "ax.legend(); ax.grid()\n",
    "ax.set_title('Accuracy');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Distribution of prediction probabilities"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2871b4bd080>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(8,5))\n",
    "prob_max = np.max(test_preds, 1)\n",
    "ax.hist(prob_max, bins=50);"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Testing accuracy on the first 1500 test elements"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Accuracy: 93.67%\n"
     ]
    }
   ],
   "source": [
    "test_acc = 100*np.mean(np.argmax(test_preds, axis=1) == y_test[:1500])\n",
    "print(\"Test Accuracy: {:0.2f}%\".format(test_acc))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}