Dropout.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Dropout\n",
    "Dropout [1] is a technique for regularizing neural networks by randomly setting some features to zero during the forward pass. In this exercise you will implement a dropout layer and modify your fully-connected network to optionally use dropout.\n",
    "\n",
    "[1] Geoffrey E. Hinton et al, \"Improving neural networks by preventing co-adaptation of feature detectors\", arXiv 2012"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The autoreload extension is already loaded. To reload it, use:\n",
      "  %reload_ext autoreload\n"
     ]
    }
   ],
   "source": [
    "# As usual, a bit of setup\n",
    "\n",
    "import time\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from cs231n.classifiers.fc_net import *\n",
    "from cs231n.data_utils import get_CIFAR10_data\n",
    "from cs231n.gradient_check import eval_numerical_gradient, eval_numerical_gradient_array\n",
    "from cs231n.solver import Solver\n",
    "\n",
    "%matplotlib inline\n",
    "plt.rcParams['figure.figsize'] = (10.0, 8.0) # set default size of plots\n",
    "plt.rcParams['image.interpolation'] = 'nearest'\n",
    "plt.rcParams['image.cmap'] = 'gray'\n",
    "\n",
    "# for auto-reloading external modules\n",
    "# see http://stackoverflow.com/questions/1907993/autoreload-of-modules-in-ipython\n",
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "def rel_error(x, y):\n",
    "  \"\"\" returns relative error \"\"\"\n",
    "  return np.max(np.abs(x - y) / (np.maximum(1e-8, np.abs(x) + np.abs(y))))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X_val:  (1000, 3, 32, 32)\n",
      "X_train:  (49000, 3, 32, 32)\n",
      "X_test:  (1000, 3, 32, 32)\n",
      "y_val:  (1000,)\n",
      "y_train:  (49000,)\n",
      "y_test:  (1000,)\n"
     ]
    }
   ],
   "source": [
    "# Load the (preprocessed) CIFAR10 data.\n",
    "\n",
    "data = get_CIFAR10_data()\n",
    "for k, v in data.iteritems():\n",
    "  print '%s: ' % k, v.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Dropout forward pass\n",
    "In the file `cs231n/layers.py`, implement the forward pass for dropout. Since dropout behaves differently during training and testing, make sure to implement the operation for both modes.\n",
    "\n",
    "Once you have done so, run the cell below to test your implementation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running tests with p =  0.3\n",
      "Mean of input:  9.99962684915\n",
      "Mean of train-time output:  7.00835310499\n",
      "Mean of test-time output:  9.99962684915\n",
      "Fraction of train-time output set to zero:  0.299136\n",
      "Fraction of test-time output set to zero:  0.0\n",
      "\n",
      "Running tests with p =  0.6\n",
      "Mean of input:  9.99962684915\n",
      "Mean of train-time output:  4.00571621862\n",
      "Mean of test-time output:  9.99962684915\n",
      "Fraction of train-time output set to zero:  0.599468\n",
      "Fraction of test-time output set to zero:  0.0\n",
      "\n",
      "Running tests with p =  0.75\n",
      "Mean of input:  9.99962684915\n",
      "Mean of train-time output:  2.48497937778\n",
      "Mean of test-time output:  9.99962684915\n",
      "Fraction of train-time output set to zero:  0.751568\n",
      "Fraction of test-time output set to zero:  0.0\n",
      "\n"
     ]
    }
   ],
   "source": [
    "x = np.random.randn(500, 500) + 10\n",
    "\n",
    "for p in [0.3, 0.6, 0.75]:\n",
    "  out, _ = dropout_forward(x, {'mode': 'train', 'p': p})\n",
    "  out_test, _ = dropout_forward(x, {'mode': 'test', 'p': p})\n",
    "\n",
    "  print 'Running tests with p = ', p\n",
    "  print 'Mean of input: ', x.mean()\n",
    "  print 'Mean of train-time output: ', out.mean()\n",
    "  print 'Mean of test-time output: ', out_test.mean()\n",
    "  print 'Fraction of train-time output set to zero: ', (out == 0).mean()\n",
    "  print 'Fraction of test-time output set to zero: ', (out_test == 0).mean()\n",
    "  print"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Dropout backward pass\n",
    "In the file `cs231n/layers.py`, implement the backward pass for dropout. After doing so, run the following cell to numerically gradient-check your implementation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dx relative error:  1.89289478629e-11\n"
     ]
    }
   ],
   "source": [
    "x = np.random.randn(10, 10) + 10\n",
    "dout = np.random.randn(*x.shape)\n",
    "\n",
    "dropout_param = {'mode': 'train', 'p': 0.8, 'seed': 123}\n",
    "out, cache = dropout_forward(x, dropout_param)\n",
    "dx = dropout_backward(dout, cache)\n",
    "dx_num = eval_numerical_gradient_array(lambda xx: dropout_forward(xx, dropout_param)[0], x, dout)\n",
    "\n",
    "print 'dx relative error: ', rel_error(dx, dx_num)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Fully-connected nets with Dropout\n",
    "In the file `cs231n/classifiers/fc_net.py`, modify your implementation to use dropout. Specificially, if the constructor the the net receives a nonzero value for the `dropout` parameter, then the net should add dropout immediately after every ReLU nonlinearity. After doing so, run the following to numerically gradient-check your implementation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running check with dropout =  0\n",
      "Initial loss:  2.31222134939\n",
      "W1 relative error: 2.35e-07\n",
      "W2 relative error: 5.52e-07\n",
      "W3 relative error: 2.83e-08\n",
      "b1 relative error: 1.89e-08\n",
      "b2 relative error: 2.01e-09\n",
      "b3 relative error: 1.95e-10\n",
      "\n",
      "Running check with dropout =  0.25\n",
      "Initial loss:  2.30220233899\n",
      "W1 relative error: 5.15e-05\n",
      "W2 relative error: 5.17e-07\n",
      "W3 relative error: 2.36e-07\n",
      "b1 relative error: 5.60e-06\n",
      "b2 relative error: 8.97e-10\n",
      "b3 relative error: 6.97e-11\n",
      "\n",
      "Running check with dropout =  0.5\n",
      "Initial loss:  2.30246506078\n",
      "W1 relative error: 5.95e-07\n",
      "W2 relative error: 5.76e-07\n",
      "W3 relative error: 7.12e-08\n",
      "b1 relative error: 5.12e-08\n",
      "b2 relative error: 1.40e-08\n",
      "b3 relative error: 8.28e-11\n",
      "\n"
     ]
    }
   ],
   "source": [
    "N, D, H1, H2, C = 2, 15, 20, 30, 10\n",
    "X = np.random.randn(N, D)\n",
    "y = np.random.randint(C, size=(N,))\n",
    "\n",
    "for dropout in [0, 0.25, 0.5]:\n",
    "  print 'Running check with dropout = ', dropout\n",
    "  model = FullyConnectedNet([H1, H2], input_dim=D, num_classes=C,\n",
    "                            weight_scale=5e-2, dtype=np.float64,\n",
    "                            dropout=dropout, seed=123)\n",
    "\n",
    "  loss, grads = model.loss(X, y)\n",
    "  print 'Initial loss: ', loss\n",
    "\n",
    "  for name in sorted(grads):\n",
    "    f = lambda _: model.loss(X, y)[0]\n",
    "    grad_num = eval_numerical_gradient(f, model.params[name], verbose=False, h=1e-5)\n",
    "    print '%s relative error: %.2e' % (name, rel_error(grad_num, grads[name]))\n",
    "  print"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Regularization experiment\n",
    "As an experiment, we will train a pair of two-layer networks on 500 training examples: one will use no dropout, and one will use a dropout probability of 0.75. We will then visualize the training and validation accuracies of the two networks over time."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "(Iteration 1 / 125) loss: 8.596245\n",
      "(Epoch 0 / 25) train acc: 0.252000; val_acc: 0.177000\n",
      "(Epoch 1 / 25) train acc: 0.284000; val_acc: 0.184000\n",
      "(Epoch 2 / 25) train acc: 0.348000; val_acc: 0.215000\n",
      "(Epoch 3 / 25) train acc: 0.424000; val_acc: 0.216000\n",
      "(Epoch 4 / 25) train acc: 0.448000; val_acc: 0.240000\n",
      "(Epoch 5 / 25) train acc: 0.534000; val_acc: 0.249000\n",
      "(Epoch 6 / 25) train acc: 0.628000; val_acc: 0.267000\n",
      "(Epoch 7 / 25) train acc: 0.678000; val_acc: 0.264000\n",
      "(Epoch 8 / 25) train acc: 0.744000; val_acc: 0.301000\n",
      "(Epoch 9 / 25) train acc: 0.792000; val_acc: 0.283000\n",
      "(Epoch 10 / 25) train acc: 0.854000; val_acc: 0.301000\n",
      "(Epoch 11 / 25) train acc: 0.846000; val_acc: 0.279000\n",
      "(Epoch 12 / 25) train acc: 0.886000; val_acc: 0.279000\n",
      "(Epoch 13 / 25) train acc: 0.916000; val_acc: 0.287000\n",
      "(Epoch 14 / 25) train acc: 0.946000; val_acc: 0.282000\n",
      "(Epoch 15 / 25) train acc: 0.960000; val_acc: 0.308000\n",
      "(Epoch 16 / 25) train acc: 0.972000; val_acc: 0.311000\n",
      "(Epoch 17 / 25) train acc: 0.968000; val_acc: 0.308000\n",
      "(Epoch 18 / 25) train acc: 0.980000; val_acc: 0.315000\n",
      "(Epoch 19 / 25) train acc: 0.968000; val_acc: 0.315000\n",
      "(Epoch 20 / 25) train acc: 0.980000; val_acc: 0.305000\n",
      "(Iteration 101 / 125) loss: 0.448314\n",
      "(Epoch 21 / 25) train acc: 0.990000; val_acc: 0.307000\n",
      "(Epoch 22 / 25) train acc: 0.986000; val_acc: 0.319000\n",
      "(Epoch 23 / 25) train acc: 0.992000; val_acc: 0.308000\n",
      "(Epoch 24 / 25) train acc: 0.994000; val_acc: 0.303000\n",
      "(Epoch 25 / 25) train acc: 0.992000; val_acc: 0.307000\n",
      "0.75\n",
      "(Iteration 1 / 125) loss: 4.501513\n",
      "(Epoch 0 / 25) train acc: 0.220000; val_acc: 0.188000\n",
      "(Epoch 1 / 25) train acc: 0.340000; val_acc: 0.242000\n",
      "(Epoch 2 / 25) train acc: 0.384000; val_acc: 0.252000\n",
      "(Epoch 3 / 25) train acc: 0.432000; val_acc: 0.260000\n",
      "(Epoch 4 / 25) train acc: 0.484000; val_acc: 0.291000\n",
      "(Epoch 5 / 25) train acc: 0.490000; val_acc: 0.265000\n",
      "(Epoch 6 / 25) train acc: 0.490000; val_acc: 0.291000\n",
      "(Epoch 7 / 25) train acc: 0.558000; val_acc: 0.275000\n",
      "(Epoch 8 / 25) train acc: 0.620000; val_acc: 0.324000\n",
      "(Epoch 9 / 25) train acc: 0.594000; val_acc: 0.282000\n",
      "(Epoch 10 / 25) train acc: 0.644000; val_acc: 0.279000\n",
      "(Epoch 11 / 25) train acc: 0.686000; val_acc: 0.314000\n",
      "(Epoch 12 / 25) train acc: 0.708000; val_acc: 0.315000\n",
      "(Epoch 13 / 25) train acc: 0.670000; val_acc: 0.298000\n",
      "(Epoch 14 / 25) train acc: 0.724000; val_acc: 0.314000\n",
      "(Epoch 15 / 25) train acc: 0.736000; val_acc: 0.281000\n",
      "(Epoch 16 / 25) train acc: 0.784000; val_acc: 0.307000\n",
      "(Epoch 17 / 25) train acc: 0.790000; val_acc: 0.301000\n",
      "(Epoch 18 / 25) train acc: 0.820000; val_acc: 0.330000\n",
      "(Epoch 19 / 25) train acc: 0.792000; val_acc: 0.316000\n",
      "(Epoch 20 / 25) train acc: 0.830000; val_acc: 0.351000\n",
      "(Iteration 101 / 125) loss: inf\n",
      "(Epoch 21 / 25) train acc: 0.840000; val_acc: 0.318000\n",
      "(Epoch 22 / 25) train acc: 0.846000; val_acc: 0.290000\n",
      "(Epoch 23 / 25) train acc: 0.858000; val_acc: 0.302000\n",
      "(Epoch 24 / 25) train acc: 0.890000; val_acc: 0.332000\n",
      "(Epoch 25 / 25) train acc: 0.896000; val_acc: 0.326000\n"
     ]
    }
   ],
   "source": [
    "# Train two identical nets, one with dropout and one without\n",
    "\n",
    "num_train = 500\n",
    "small_data = {\n",
    "  'X_train': data['X_train'][:num_train],\n",
    "  'y_train': data['y_train'][:num_train],\n",
    "  'X_val': data['X_val'],\n",
    "  'y_val': data['y_val'],\n",
    "}\n",
    "\n",
    "solvers = {}\n",
    "dropout_choices = [0, 0.75]\n",
    "for dropout in dropout_choices:\n",
    "  model = FullyConnectedNet([500], dropout=dropout)\n",
    "  print dropout\n",
    "\n",
    "  solver = Solver(model, small_data,\n",
    "                  num_epochs=25, batch_size=100,\n",
    "                  update_rule='adam',\n",
    "                  optim_config={\n",
    "                    'learning_rate': 5e-4,\n",
    "                  },\n",
    "                  verbose=True, print_every=100)\n",
    "  solver.train()\n",
    "  solvers[dropout] = solver"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA4MAAAJXCAYAAAApT4F3AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nd1+hd2tTEwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f00e031c950>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot train and validation accuracies of the two models\n",
    "\n",
    "train_accs = []\n",
    "val_accs = []\n",
    "for dropout in dropout_choices:\n",
    "  solver = solvers[dropout]\n",
    "  train_accs.append(solver.train_acc_history[-1])\n",
    "  val_accs.append(solver.val_acc_history[-1])\n",
    "\n",
    "plt.subplot(3, 1, 1)\n",
    "for dropout in dropout_choices:\n",
    "  plt.plot(solvers[dropout].train_acc_history, 'o', label='%.2f dropout' % dropout)\n",
    "plt.title('Train accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend(ncol=2, loc='lower right')\n",
    "  \n",
    "plt.subplot(3, 1, 2)\n",
    "for dropout in dropout_choices:\n",
    "  plt.plot(solvers[dropout].val_acc_history, 'o', label='%.2f dropout' % dropout)\n",
    "plt.title('Val accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend(ncol=2, loc='lower right')\n",
    "\n",
    "plt.gcf().set_size_inches(15, 15)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Question\n",
    "Explain what you see in this experiment. What does it suggest about dropout?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Answer\n",
    "Obtaining very similar validation results the model with dropout had poorer training results, but closer to the outcome in a \"real\" situation. Hence, with the dropout the overfitting decreased.\n",
    "\n",
    "In a similar way to regularization, dropouts make the net learn without relying on specific values of its nodes as they are masked arbitrarily, and make it pay attention to the combination of them. Thus, it avoid to overfit on specific nodes and values.\n",
    "\n",
    "Also, see that this configuration naturally overfits due to the relation in size between the data and parameters."
   ]
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "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": "2.7.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}