corrnet.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Correlational Neural Networks with Tensorflow\n",
    "\n",
    "![alt text](https://deeplearnschool.files.wordpress.com/2017/05/cornet.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Import Modules\n",
    "\n",
    "Import all the necessary modules for training and testing the CorrNet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Populating the interactive namespace from numpy and matplotlib\n"
     ]
    }
   ],
   "source": [
    "%pylab inline\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "import tensorflow.contrib.keras as keras\n",
    "\n",
    "# Our modules\n",
    "import corrnet.data.mnist_views\n",
    "import corrnet.layers\n",
    "import corrnet.loss"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Fetch and split MNIST data\n",
    "\n",
    "left_mnist contains the left halves of images and right_mnist contains the right halves\n",
    "Usage:\n",
    "- left_mnist.train.images: Left images for training\n",
    "- left_mnist.validation.images: Left images for validation\n",
    "- left_mnist.test.images: Left images for testing\n",
    "\n",
    "Example of a MNIST image left and right halves:\n",
    "![alt text](https://deeplearnschool.files.wordpress.com/2017/05/d2.png?w=340&h=245)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "left_mnist = corrnet.data.mnist_views.left_view_dataset()\n",
    "right_mnist = corrnet.data.mnist_views.right_view_dataset()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# CorrNet and training configuration"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "n_inputs = 28*14\n",
    "hdim1, hdim2, hdim = 500, 300, 50\n",
    "learning_rate = 0.001\n",
    "n_epochs = 40\n",
    "batch_size = 100\n",
    "lamda = 0.02\n",
    "\n",
    "tf.reset_default_graph()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# Build CorrNet with tensorflow\n",
    "\n",
    "We will build a NN with 3 encoding layers (for both modalities), 1 common layer and 1 decoding layer (for both modalities).\n",
    "\n",
    "- The encoding layers will build a \"pyramid\" with decreasing hidden units (l1: 500 -> l2: 300 -> l3: 50)\n",
    "- The common layer adds the output of l3 for left and right input"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X1 = tf.placeholder(tf.float32, shape=(None,n_inputs), name='X1')\n",
    "X2 = tf.placeholder(tf.float32, shape=(None, n_inputs), name='X2')\n",
    "keep_prob = tf.placeholder(tf.float32)\n",
    "\n",
    "with tf.name_scope('dnn'):\n",
    "    h1_X1_l = corrnet.layers.dense_layer(X1, hdim1, name='h1_X1', activation=corrnet.activations.leaky_relu)\n",
    "    h1_o = h1_X1_l.apply(X1)\n",
    "    h1_X2_l = corrnet.layers.dense_layer(X2, hdim1, name='h1_X2', activation=corrnet.activations.leaky_relu)\n",
    "    h2_o = h1_X2_l.apply(X2)\n",
    "\n",
    "    h2_X1_l = corrnet.layers.dense_layer(h1_o, hdim2, name='h2_X1', activation=corrnet.activations.leaky_relu)\n",
    "    h1_o = h2_X1_l.apply(h1_o)\n",
    "    h2_X2_l = corrnet.layers.dense_layer(h2_o, hdim2, name='h2_X2', activation=corrnet.activations.leaky_relu)\n",
    "    h2_o = h2_X2_l.apply(h2_o)\n",
    "\n",
    "    h1_l = corrnet.layers.dense_layer(h1_o, hdim, name='h1', activation=corrnet.activations.leaky_relu)\n",
    "    h1_o = h1_l(h1_o)\n",
    "    h2_l = corrnet.layers.dense_layer(h2_o, hdim, name='h2', activation=corrnet.activations.leaky_relu)\n",
    "    h2_o = h2_l(h2_o)\n",
    "    h_l = keras.layers.Add()\n",
    "    h_o = h_l([h1_o, h2_o])\n",
    "    \n",
    "    r1_X1_l = corrnet.layers.dense_layer(h_o, n_inputs, name='r1_X1', activation=corrnet.activations.leaky_relu)\n",
    "    reconstruct_X1 = r1_X1_l.apply(h_o)\n",
    "\n",
    "    r1_X2_l = corrnet.layers.dense_layer(h_o, n_inputs, name='r1_X2', activation=corrnet.activations.leaky_relu)\n",
    "    reconstruct_X2 = r1_X2_l.apply(h_o)    \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# Define loss, optimizer and evaluation metric\n",
    "\n",
    "The loss function is defined as:\n",
    "![alt text](https://deeplearnschool.files.wordpress.com/2017/05/eq31.png?w=640)\n",
    "\n",
    "- We use all terms and set lambda=0.02\n",
    "- Adam optimizer with learning_rate=0.001 ensures fast and stable convergence"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "with tf.name_scope('loss'):\n",
    "    loss = corrnet.loss.calculate_loss(\n",
    "        [h1_X1_l, h2_X1_l, h1_l],\n",
    "        [h1_X2_l, h2_X2_l, h2_l],\n",
    "        [r1_X1_l], [r1_X2_l],\n",
    "        h_l, [X1, X2],\n",
    "        ['l1', 'l2', 'l3', 'l4'], lamda=lamda)\n",
    "\n",
    "with tf.name_scope('train'):\n",
    "    optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate)\n",
    "    train_op = optimizer.minimize(loss)\n",
    "\n",
    "with tf.name_scope('eval'):\n",
    "    reconstruction_accuracy = loss"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# Train model\n",
    "\n",
    "Use minibatch training with batch_size = 100 for 40 epochs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 0\tTrain accuracy: 0.135617911816\tValidation accuracy: 7.59412050247\n",
      "Epoch: 1\tTrain accuracy: 0.105549052358\tValidation accuracy: 6.4984254837\n",
      "Epoch: 2\tTrain accuracy: 0.0904521644115\tValidation accuracy: 5.69390630722\n",
      "Epoch: 3\tTrain accuracy: 0.0826511234045\tValidation accuracy: 5.26948547363\n",
      "Epoch: 4\tTrain accuracy: 0.074488542974\tValidation accuracy: 4.91267156601\n",
      "Epoch: 5\tTrain accuracy: 0.0699452906847\tValidation accuracy: 4.60951948166\n",
      "Epoch: 6\tTrain accuracy: 0.0665562897921\tValidation accuracy: 4.41534280777\n",
      "Epoch: 7\tTrain accuracy: 0.0635370761156\tValidation accuracy: 4.23102045059\n",
      "Epoch: 8\tTrain accuracy: 0.0608576983213\tValidation accuracy: 4.13588523865\n",
      "Epoch: 9\tTrain accuracy: 0.0599887259305\tValidation accuracy: 4.07832813263\n",
      "Epoch: 10\tTrain accuracy: 0.0573939830065\tValidation accuracy: 3.94658398628\n",
      "Epoch: 11\tTrain accuracy: 0.0561895817518\tValidation accuracy: 3.91816234589\n",
      "Epoch: 12\tTrain accuracy: 0.0548528395593\tValidation accuracy: 3.91364789009\n",
      "Epoch: 13\tTrain accuracy: 0.0545355677605\tValidation accuracy: 3.92132592201\n",
      "Epoch: 14\tTrain accuracy: 0.052734836936\tValidation accuracy: 3.87472629547\n",
      "Epoch: 15\tTrain accuracy: 0.0510214418173\tValidation accuracy: 3.83363604546\n",
      "Epoch: 16\tTrain accuracy: 0.0500749647617\tValidation accuracy: 3.7941596508\n",
      "Epoch: 17\tTrain accuracy: 0.0487093403935\tValidation accuracy: 3.78180122375\n",
      "Epoch: 18\tTrain accuracy: 0.0477666184306\tValidation accuracy: 3.72977256775\n",
      "Epoch: 19\tTrain accuracy: 0.0484120175242\tValidation accuracy: 3.81505036354\n",
      "Epoch: 20\tTrain accuracy: 0.0485107451677\tValidation accuracy: 3.81252479553\n",
      "Epoch: 21\tTrain accuracy: 0.0482228398323\tValidation accuracy: 3.81815958023\n",
      "Epoch: 22\tTrain accuracy: 0.0488468296826\tValidation accuracy: 3.85982465744\n",
      "Epoch: 23\tTrain accuracy: 0.0477372035384\tValidation accuracy: 3.79056954384\n",
      "Epoch: 24\tTrain accuracy: 0.0457429625094\tValidation accuracy: 3.64555573463\n",
      "Epoch: 25\tTrain accuracy: 0.0445291809738\tValidation accuracy: 3.59961748123\n",
      "Epoch: 26\tTrain accuracy: 0.0453226342797\tValidation accuracy: 3.66246795654\n",
      "Epoch: 27\tTrain accuracy: 0.0448046848178\tValidation accuracy: 3.63453102112\n",
      "Epoch: 28\tTrain accuracy: 0.0442233420908\tValidation accuracy: 3.62417888641\n",
      "Epoch: 29\tTrain accuracy: 0.0441804602742\tValidation accuracy: 3.59581685066\n",
      "Epoch: 30\tTrain accuracy: 0.041626766324\tValidation accuracy: 3.49308919907\n",
      "Epoch: 31\tTrain accuracy: 0.0405859164894\tValidation accuracy: 3.49122285843\n",
      "Epoch: 32\tTrain accuracy: 0.0402261987329\tValidation accuracy: 3.49294304848\n",
      "Epoch: 33\tTrain accuracy: 0.0403299480677\tValidation accuracy: 3.47547674179\n",
      "Epoch: 34\tTrain accuracy: 0.0404042229056\tValidation accuracy: 3.4726023674\n",
      "Epoch: 35\tTrain accuracy: 0.0411220118403\tValidation accuracy: 3.54630088806\n",
      "Epoch: 36\tTrain accuracy: 0.0400619283319\tValidation accuracy: 3.48062229156\n",
      "Epoch: 37\tTrain accuracy: 0.0944200009108\tValidation accuracy: 6.40904808044\n",
      "Epoch: 38\tTrain accuracy: 0.0400039739907\tValidation accuracy: 3.4560508728\n",
      "Epoch: 39\tTrain accuracy: 0.0399003848433\tValidation accuracy: 3.47544288635\n",
      "Test accuracy: 7.07056808472\n"
     ]
    }
   ],
   "source": [
    "init = tf.global_variables_initializer()\n",
    "saver = tf.train.Saver()\n",
    "\n",
    "with tf.Session() as sess:\n",
    "    init.run()\n",
    "    for epoch in range(n_epochs):\n",
    "        for iteration in range(left_mnist.train.num_examples // batch_size):\n",
    "            X_batch1, _ = left_mnist.train.next_batch(batch_size, shuffle=False)\n",
    "            X_batch2, _ = right_mnist.train.next_batch(batch_size, shuffle=False)\n",
    "            sess.run(train_op, feed_dict={X1: X_batch1, X2: X_batch2})\n",
    "        acc_train = reconstruction_accuracy.eval(feed_dict={X1: X_batch1, X2: X_batch2})\n",
    "        X_val_batch1 = left_mnist.validation.images\n",
    "        X_val_batch2 = right_mnist.validation.images\n",
    "        acc_validation = reconstruction_accuracy.eval(feed_dict={X1: X_val_batch1, X2: X_val_batch2})\n",
    "        print('Epoch: {}\\tTrain accuracy: {}\\tValidation accuracy: {}'.format(epoch, acc_train, acc_validation))\n",
    "\n",
    "    X_test1 = left_mnist.test.images\n",
    "    X_test2 = right_mnist.test.images\n",
    "    rec_left, rec_right = sess.run([reconstruct_X1, reconstruct_X2], feed_dict={X1: X_test1, X2: X_test2})\n",
    "    rec_left_left, rec_left_right = sess.run([reconstruct_X1, reconstruct_X2], feed_dict={X1: X_test1, X2: np.zeros_like(X_test2)})\n",
    "    rec_right_left, rec_right_right = sess.run([reconstruct_X1, reconstruct_X2], feed_dict={X1: np.zeros_like(X_test1), X2: X_test2})\n",
    "\n",
    "    acc_test = reconstruction_accuracy.eval(feed_dict={X1: X_test1, X2: X_test2})\n",
    "    print('Test accuracy: {}'.format(acc_test))\n",
    "    save_path = saver.save(sess, './my_model_final_multi.ckpt')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Visual Evaluation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Original example image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7efcbaefab50>"
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAcwAAAEICAYAAADSosQkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nEJgAABQgMAEAKEBgAgBQ4P8DRBPKOFgUhEMAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7efcbc169a50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "f, (ax1, ax2) = plt.subplots(1, 2, sharey=True)\n",
    "ax1.imshow(np.reshape(left_mnist.test.images[4], (28,14)))\n",
    "ax1.set_title('The original MNIST left and right halves for an unseen image')\n",
    "ax2.imshow(np.reshape(right_mnist.test.images[4], (28,14)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Reconstruction when both and right images are provided to the NN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7efca87f6ed0>"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAhcAAAEICAYAAADssdabAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n1VUxNwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7efca8f79ad0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "f, (ax1, ax2) = plt.subplots(1, 2, sharey=True)\n",
    "ax1.imshow(np.reshape(rec_left[4], (28,14)))\n",
    "ax1.set_title('Reconstructed left and right image when both left and right halves are given as input')\n",
    "ax2.imshow(np.reshape(rec_right[4], (28,14)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Self and cross reconstruction when only left image is provided"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7efcb0167810>"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAesAAAEICAYAAAB/I9yNAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n4ZSsRURECqdkLSIiUrj/D1E7zE+PkdMSAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7efca8bc3ad0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "f, (ax1, ax2) = plt.subplots(1, 2, sharey=True)\n",
    "ax1.imshow(np.reshape(rec_left_left[4], (28,14)))\n",
    "ax1.set_title('Reconstructed left and right image when only left half is given as input')\n",
    "ax2.imshow(np.reshape(rec_left_right[4], (28,14)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Self and cross reconstruction when only right image is provided"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7efcbcb16f90>"
      ]
     },
     "execution_count": 67,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAe8AAAEICAYAAAB2yHz3AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nrTco4N6dAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7efca8796c50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "f, (ax1, ax2) = plt.subplots(1, 2, sharey=True)\n",
    "ax1.imshow(np.reshape(rec_right_left[4], (28,14)))\n",
    "ax1.set_title('Reconstructed left and right image when only right half is given as input')\n",
    "ax2.imshow(np.reshape(rec_right_right[4], (28,14)))"
   ]
  }
 ],
 "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": 1
}