modified_mnist.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## A) For MNIST dataset, run a TF in supervised mode (train/test) and report results "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Using GPU:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.6.0\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "print(tf.__version__)\n",
    "tf.logging.set_verbosity(tf.logging.INFO)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "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": [
    "#download mnist data\n",
    "from tensorflow.examples.tutorials.mnist import input_data\n",
    "mnist = input_data.read_data_sets('MNIST_data', one_hot=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "#start interactive session\n",
    "sess = tf.InteractiveSession(config=tf.ConfigProto(log_device_placement=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#data exploration\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "plt.imshow(np.reshape(mnist.train.images[0], [28, 28]), cmap='gray') #current shape is 28 x 28\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Drop classes 8 and 9:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "mnist_train_data = np.array([x for (x,y) in zip(mnist.train.images,mnist.train.labels) if y[8]==0 and y[9]==0])\n",
    "mnist_train_labels = np.array([y[0:8] for y in mnist.train.labels if y[8]==0 and y[9]==0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((44157, 784), (44157, 8))"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mnist_train_data.shape, mnist_train_labels.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "mnist_test_data = np.array([x for (x,y) in zip(mnist.test.images,mnist.test.labels) if y[8]==0 and y[9]==0])\n",
    "mnist_test_labels = np.array([y[0:8] for y in mnist.test.labels if y[8]==0 and y[9]==0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((8017, 784), (8017, 8))"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mnist_test_data.shape, mnist_test_labels.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### resize_by_padding zeroes to 28x28 image to make it 32x32 "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "def resize_by_padding(images):\n",
    "    resized_images = []\n",
    "    print(\"Total\", len(images), \"images..\")\n",
    "    \n",
    "    for i in range(len(images)):\n",
    "        \n",
    "        if i%1000 == 0:\n",
    "            print(\"Resized\", i, \"images till now...\")\n",
    "        \n",
    "        \"\"\"\n",
    "        t= tf.constant(images[i], shape=[28,28])\n",
    "        paddings=tf.constant([[2,2], [2,2]])\n",
    "        var = tf.pad(t, paddings, \"CONSTANT\")  #converting to array\n",
    "        b = tf.reshape(var,[1024]) \n",
    "        tempnmpy = b.eval()\n",
    "        \"\"\"\n",
    "        tempnmpy = np.zeros([32,32])\n",
    "        b = np.reshape(images[i], [28, 28])\n",
    "        tempnmpy[:b.shape[0], :b.shape[1]] = b\n",
    "        tempnmpy = tempnmpy.reshape(1024)\n",
    "        #print(tempnmpy)\n",
    "        #plt.imshow(np.reshape(tempnmpy, [32, 32]), cmap='gray')\n",
    "        #plt.show()\n",
    "        \n",
    "        resized_images.append(tempnmpy)\n",
    "    return np.array(resized_images)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total 44157 images..\n",
      "Resized 0 images till now...\n",
      "Resized 1000 images till now...\n",
      "Resized 2000 images till now...\n",
      "Resized 3000 images till now...\n",
      "Resized 4000 images till now...\n",
      "Resized 5000 images till now...\n",
      "Resized 6000 images till now...\n",
      "Resized 7000 images till now...\n",
      "Resized 8000 images till now...\n",
      "Resized 9000 images till now...\n",
      "Resized 10000 images till now...\n",
      "Resized 11000 images till now...\n",
      "Resized 12000 images till now...\n",
      "Resized 13000 images till now...\n",
      "Resized 14000 images till now...\n",
      "Resized 15000 images till now...\n",
      "Resized 16000 images till now...\n",
      "Resized 17000 images till now...\n",
      "Resized 18000 images till now...\n",
      "Resized 19000 images till now...\n",
      "Resized 20000 images till now...\n",
      "Resized 21000 images till now...\n",
      "Resized 22000 images till now...\n",
      "Resized 23000 images till now...\n",
      "Resized 24000 images till now...\n",
      "Resized 25000 images till now...\n",
      "Resized 26000 images till now...\n",
      "Resized 27000 images till now...\n",
      "Resized 28000 images till now...\n",
      "Resized 29000 images till now...\n",
      "Resized 30000 images till now...\n",
      "Resized 31000 images till now...\n",
      "Resized 32000 images till now...\n",
      "Resized 33000 images till now...\n",
      "Resized 34000 images till now...\n",
      "Resized 35000 images till now...\n",
      "Resized 36000 images till now...\n",
      "Resized 37000 images till now...\n",
      "Resized 38000 images till now...\n",
      "Resized 39000 images till now...\n",
      "Resized 40000 images till now...\n",
      "Resized 41000 images till now...\n",
      "Resized 42000 images till now...\n",
      "Resized 43000 images till now...\n",
      "Resized 44000 images till now...\n"
     ]
    }
   ],
   "source": [
    "mnist_train_data = resize_by_padding(mnist_train_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total 8017 images..\n",
      "Resized 0 images till now...\n",
      "Resized 1000 images till now...\n",
      "Resized 2000 images till now...\n",
      "Resized 3000 images till now...\n",
      "Resized 4000 images till now...\n",
      "Resized 5000 images till now...\n",
      "Resized 6000 images till now...\n",
      "Resized 7000 images till now...\n",
      "Resized 8000 images till now...\n"
     ]
    }
   ],
   "source": [
    "mnist_test_data = resize_by_padding(mnist_test_data)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "New shapes:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((44157, 1024), (44157, 8), (8017, 1024), (8017, 8))"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mnist_train_data.shape, mnist_train_labels.shape, mnist_test_data.shape, mnist_test_labels.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### All variables converted to float16"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "x = tf.placeholder(tf.float16, shape=[None, 1024])\n",
    "y_ = tf.placeholder(tf.float16, shape=[None, 8])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "W = tf.Variable(tf.zeros([1024,8], dtype=tf.float16), dtype=tf.float16)\n",
    "b = tf.Variable(tf.zeros([8], dtype=tf.float16), dtype=tf.float16)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "#run session\n",
    "sess.run(tf.global_variables_initializer())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "y = tf.matmul(x,W) + b"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From <ipython-input-21-78b4c0e0d208>:2: softmax_cross_entropy_with_logits (from tensorflow.python.ops.nn_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "\n",
      "Future major versions of TensorFlow will allow gradients to flow\n",
      "into the labels input on backprop by default.\n",
      "\n",
      "See tf.nn.softmax_cross_entropy_with_logits_v2.\n",
      "\n"
     ]
    }
   ],
   "source": [
    "cross_entropy = tf.reduce_mean(\n",
    "    tf.nn.softmax_cross_entropy_with_logits(labels=y_, logits=y))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_step = tf.train.GradientDescentOptimizer(0.5).minimize(cross_entropy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "#get batches for training\n",
    "def get_batch(X, y, batch_size = 100):\n",
    "    n = X.shape[0]\n",
    "    batch = np.floor(np.random.rand(batch_size)*n).astype(int)\n",
    "    batch_x = X[batch,:]\n",
    "    batch_y = y[batch,:]\n",
    "    return batch_x, batch_y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "#train:\n",
    "for _ in range(1000):\n",
    "    batch_x, batch_y = get_batch(mnist_train_data, mnist_train_labels, 100)\n",
    "    train_step.run(feed_dict={x: batch_x, y_: batch_y})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "correct_prediction = tf.equal(tf.argmax(y,1), tf.argmax(y_,1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float16))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.947\n"
     ]
    }
   ],
   "source": [
    "#Accuracy\n",
    "print(sess.run(accuracy, feed_dict={x: mnist_test_data, y_: mnist_test_labels}))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Predict single images\n",
    "n_images = 4\n",
    "# Get images from test set\n",
    "test_images = mnist_test_data[:n_images]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_labels = mnist_test_labels[:n_images]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prepare the input data\n",
    "input_fn = tf.estimator.inputs.numpy_input_fn(\n",
    "    x={'images': test_images}, shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "prediction=tf.argmax(y,1) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "preds = prediction.eval(feed_dict={x: test_images}, session=sess)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model prediction: 7\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model prediction: 2\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model prediction: 1\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model prediction: 0\n"
     ]
    }
   ],
   "source": [
    "# Display\n",
    "for i in range(n_images):\n",
    "    plt.imshow(np.reshape(test_images[i], [32, 32]), cmap='gray')\n",
    "    plt.show()\n",
    "    print(\"Model prediction:\", preds[i])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}