CNN_Optimizer.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting MNIST_data/train-images-idx3-ubyte.gz\n",
      "Extracting MNIST_data/train-labels-idx1-ubyte.gz\n",
      "Extracting MNIST_data/t10k-images-idx3-ubyte.gz\n",
      "Extracting MNIST_data/t10k-labels-idx1-ubyte.gz\n"
     ]
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import tensorflow as tf\n",
    "from tensorflow.examples.tutorials.mnist import input_data\n",
    "\n",
    "# Import data\n",
    "mnist = input_data.read_data_sets(\"MNIST_data/\", one_hot=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Helper functions for creating weight variables\n",
    "def weight_variable(shape):\n",
    "    \"\"\"weight_variable generates a weight variable of a given shape.\"\"\"\n",
    "    initial = tf.truncated_normal(shape, stddev=0.1)\n",
    "    return tf.Variable(initial)\n",
    "\n",
    "def bias_variable(shape):\n",
    "    \"\"\"bias_variable generates a bias variable of a given shape.\"\"\"\n",
    "    initial = tf.constant(0.1, shape=shape)\n",
    "    return tf.Variable(initial)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Convolutional neural network functions\n",
    "def conv2d(x, W):\n",
    "    \"\"\"conv2d returns a 2d convolution layer with full stride.\"\"\"\n",
    "    return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME')\n",
    "\n",
    "def max_pool_2x2(x):\n",
    "    \"\"\"max_pool_2x2 downsamples a feature map by 2X.\"\"\"\n",
    "    return tf.nn.max_pool(x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Model Inputs\n",
    "x = tf.placeholder(tf.float32, [None, 784])### MNIST images enter graph here ###\n",
    "y_ = tf.placeholder(tf.float32, [None, 10])### MNIST labels enter graph here ###\n",
    "\n",
    "# Define the graph\n",
    "x_2d = tf.reshape(x,[-1,28,28,1])\n",
    "#[filter_height, filter_width, in_channels, out_channels]\n",
    "W_conv1 = weight_variable([5, 5, 1, 32])\n",
    "b_conv1 = bias_variable([32])\n",
    "conv1_out = conv2d(x_2d,W_conv1) + b_conv1\n",
    "relu1_out = tf.nn.relu(conv1_out)\n",
    "pool1_out = max_pool_2x2(relu1_out)\n",
    "W_conv2 = weight_variable([5,5,32,64])\n",
    "b_conv2 = bias_variable([64])\n",
    "conv2_out =conv2d(pool1_out,W_conv2) + b_conv2\n",
    "relu2_out = tf.nn.relu(conv2_out)\n",
    "pool2_out = max_pool_2x2(relu2_out)\n",
    "#flatten for the FC layer\n",
    "fc1_in = tf.reshape(pool2_out,[-1,64*49])\n",
    "W_fc1 = weight_variable([64*49,1024])\n",
    "b_fc1 = bias_variable([1024])\n",
    "fc1_out = tf.nn.relu(tf.matmul(fc1_in,W_fc1) + b_fc1)\n",
    "W_fc2 = weight_variable([1024,10])\n",
    "b_fc2 = bias_variable([10])\n",
    "fc2_out = tf.matmul(fc1_out,W_fc2) + b_fc2\n",
    "y_conv = fc2_out\n",
    "\n",
    "# Loss\n",
    "cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y_, logits=y_conv))\n",
    "\n",
    "# Evaluation\n",
    "correct_prediction = tf.equal(tf.argmax(y_conv, 1), tf.argmax(y_, 1))\n",
    "accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Optimizers: Try out a few different parameters for SGD and SGD momentum\n",
    "train_step_SGD = tf.train.GradientDescentOptimizer(learning_rate=0.001).minimize(cross_entropy)\n",
    "train_step_momentum = tf.train.MomentumOptimizer(learning_rate=0.001, momentum=0.9).minimize(cross_entropy)\n",
    "train_step_ADAM = tf.train.AdamOptimizer().minimize(cross_entropy)\n",
    "\n",
    "# Op for initializing all variables\n",
    "initialize_all = tf.global_variables_initializer()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def train_CNN(train_step_optimizer, iterations=4000):\n",
    "    with tf.Session() as sess:\n",
    "        # Initialize (or reset) all variables\n",
    "        sess.run(initialize_all)\n",
    "        \n",
    "        # Initialize arrays to track losses and validation accuracies\n",
    "        valid_accs = [] \n",
    "        losses = []\n",
    "        \n",
    "        for i in range(iterations):\n",
    "            # Validate every 250th batch\n",
    "            if i % 250 == 0:\n",
    "                validation_accuracy = 0\n",
    "                for v in range(10):\n",
    "                    batch = mnist.validation.next_batch(50)\n",
    "                    validation_accuracy += (1/10) * accuracy.eval(feed_dict={x: batch[0], y_: batch[1]})\n",
    "                print('step %d, validation accuracy %g' % (i, validation_accuracy))\n",
    "                valid_accs.append(validation_accuracy)\n",
    "                \n",
    "            # Train    \n",
    "            batch = mnist.train.next_batch(50)\n",
    "            loss, _ = sess.run([cross_entropy, train_step_optimizer], feed_dict={x: batch[0], y_: batch[1]})\n",
    "            losses.append(loss)\n",
    "            \n",
    "        print('test accuracy %g' % accuracy.eval(feed_dict={x: mnist.test.images, y_: mnist.test.labels}))\n",
    "        \n",
    "    return valid_accs, losses"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "SGD:\n",
      "step 0, validation accuracy 0.092\n",
      "step 250, validation accuracy 0.788\n",
      "step 500, validation accuracy 0.876\n",
      "step 750, validation accuracy 0.886\n",
      "step 1000, validation accuracy 0.882\n",
      "step 1250, validation accuracy 0.924\n",
      "step 1500, validation accuracy 0.934\n",
      "step 1750, validation accuracy 0.942\n",
      "step 2000, validation accuracy 0.938\n",
      "step 2250, validation accuracy 0.94\n",
      "step 2500, validation accuracy 0.938\n",
      "step 2750, validation accuracy 0.934\n",
      "step 3000, validation accuracy 0.926\n",
      "step 3250, validation accuracy 0.934\n",
      "step 3500, validation accuracy 0.944\n",
      "step 3750, validation accuracy 0.944\n",
      "test accuracy 0.9518\n",
      "Momentum:\n",
      "step 0, validation accuracy 0.126\n",
      "step 250, validation accuracy 0.904\n",
      "step 500, validation accuracy 0.964\n",
      "step 750, validation accuracy 0.964\n",
      "step 1000, validation accuracy 0.952\n",
      "step 1250, validation accuracy 0.97\n",
      "step 1500, validation accuracy 0.982\n",
      "step 1750, validation accuracy 0.972\n",
      "step 2000, validation accuracy 0.97\n",
      "step 2250, validation accuracy 0.982\n",
      "step 2500, validation accuracy 0.976\n",
      "step 2750, validation accuracy 0.976\n",
      "step 3000, validation accuracy 0.972\n",
      "step 3250, validation accuracy 0.982\n",
      "step 3500, validation accuracy 0.972\n",
      "step 3750, validation accuracy 0.984\n",
      "test accuracy 0.9811\n",
      "ADAM:\n",
      "step 0, validation accuracy 0.092\n",
      "step 250, validation accuracy 0.978\n",
      "step 500, validation accuracy 0.972\n",
      "step 750, validation accuracy 0.992\n",
      "step 1000, validation accuracy 0.986\n",
      "step 1250, validation accuracy 0.98\n",
      "step 1500, validation accuracy 0.988\n",
      "step 1750, validation accuracy 0.982\n",
      "step 2000, validation accuracy 0.986\n",
      "step 2250, validation accuracy 0.986\n",
      "step 2500, validation accuracy 0.99\n",
      "step 2750, validation accuracy 0.988\n",
      "step 3000, validation accuracy 0.98\n",
      "step 3250, validation accuracy 0.998\n",
      "step 3500, validation accuracy 0.994\n",
      "step 3750, validation accuracy 0.984\n",
      "test accuracy 0.9873\n"
     ]
    }
   ],
   "source": [
    "print(\"SGD:\")\n",
    "valid_accs_SGD, losses_SGD = train_CNN(train_step_SGD)\n",
    "print(\"Momentum:\")\n",
    "valid_accs_momentum, losses_momentum = train_CNN(train_step_momentum)\n",
    "print(\"ADAM:\")\n",
    "valid_accs_ADAM, losses_ADAM = train_CNN(train_step_ADAM)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x151022f7c50>"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAbMAAAEYCAYAAADWNhiqAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nOzONRqPRNHu0M9NoNBpNs+f/AdrGTeQls32CAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x151006b8128>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(1, 2)\n",
    "fig.tight_layout()\n",
    "\n",
    "ax[0].plot(valid_accs_SGD)\n",
    "ax[0].plot(valid_accs_momentum)\n",
    "ax[0].plot(valid_accs_ADAM)\n",
    "\n",
    "ax[0].set_ylabel('Validation Accuracy')\n",
    "ax[0].legend(['SGD', 'Momentum', 'ADAM'], loc='lower right')\n",
    "\n",
    "ax[1].plot(losses_SGD)\n",
    "ax[1].plot(losses_momentum)\n",
    "ax[1].plot(losses_ADAM)\n",
    "\n",
    "ax[1].set_ylabel('Cross Entropy')\n",
    "ax[1].legend(['SGD', 'Momentum', 'ADAM'], loc='upper right')\n",
    "# ax[1].set_ylim([0,1.5])  # <- Use this to change y-axis limits"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "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": "3.6.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}