Mnist with Tensorflow.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\vas\\AppData\\Local\\Continuum\\anaconda3\\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 matplotlib.pyplot as plt\n",
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "from sklearn.metrics import confusion_matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting data/MNIST/train-images-idx3-ubyte.gz\n",
      "Extracting data/MNIST/train-labels-idx1-ubyte.gz\n",
      "Extracting data/MNIST/t10k-images-idx3-ubyte.gz\n",
      "Extracting data/MNIST/t10k-labels-idx1-ubyte.gz\n"
     ]
    }
   ],
   "source": [
    "from tensorflow.examples.tutorials.mnist import input_data\n",
    "data = input_data.read_data_sets('data/MNIST/', one_hot=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0., 0., 0., 0., 0., 0., 0., 1., 0., 0.],\n",
       "       [0., 0., 1., 0., 0., 0., 0., 0., 0., 0.],\n",
       "       [0., 1., 0., 0., 0., 0., 0., 0., 0., 0.],\n",
       "       [1., 0., 0., 0., 0., 0., 0., 0., 0., 0.],\n",
       "       [0., 0., 0., 0., 1., 0., 0., 0., 0., 0.]])"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.test.labels[0:5, :]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([7, 2, 1, 0, 4], dtype=int64)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.test.cls = np.array([label.argmax() for label in data.test.labels])\n",
    "data.test.cls[0:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "from argparse import Namespace\n",
    "args = Namespace(img_size = 28,\n",
    "    img_size_flat = 28 * 28,\n",
    "    img_shape = (28, 28),\n",
    "    num_classes = 10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_images(images, cls_true, cls_pred=None):\n",
    "    assert len(images) == len(cls_true) == 9\n",
    "    \n",
    "    fig, axes = plt.subplots(3, 3)\n",
    "    fig.subplots_adjust(hspace=0.3, wspace=0)\n",
    "    \n",
    "    for i, ax in enumerate(axes.flat):\n",
    "        ax.imshow(images[i].reshape(args.img_shape), cmap=plt.cm.plasma)\n",
    "        \n",
    "        if cls_pred is None:\n",
    "            xlabel = 'True: {0}'.format(cls_true[i])\n",
    "        else:\n",
    "            xlabel = 'True: {0}, Pred: {1}'.format(cls_true[i], cls_pred[i])\n",
    "        ax.set_xlabel(xlabel)\n",
    "        ax.set_xticks([])\n",
    "        ax.set_yticks([])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 9 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "images = data.test.images [0 : 9]\n",
    "cls_true = data.test.cls [0 : 9]\n",
    "plot_images(images=images, cls_true=cls_true)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "x = tf.placeholder(tf.float32, [None, args.img_size_flat])\n",
    "y_true = tf.placeholder(tf.float32, [None, args.num_classes])\n",
    "y_true_cls = tf.placeholder(tf.int64, [None])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "weights = tf.Variable(tf.zeros([args.img_size_flat, args.num_classes]))\n",
    "biases = tf.Variable(tf.zeros([args.num_classes]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "logit = tf.matmul(x, weights) + biases"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = tf.nn.softmax(logit)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred_cls = tf.argmax(y_pred, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "cross_entropy = tf.nn.softmax_cross_entropy_with_logits_v2(logits=logit, labels=y_true)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "cost = tf.reduce_mean(cross_entropy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.3).minimize(cost)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "correct_prediction = tf.equal(y_pred_cls, y_true_cls)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "session = tf.Session()\n",
    "session.run(tf.global_variables_initializer())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch_size = 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "def optimize(num_iterations):\n",
    "    for i in range(num_iterations):\n",
    "        x_batch, y_true_batch = data.train.next_batch(batch_size=batch_size)\n",
    "        feed_dict_train = {x: x_batch, y_true: y_true_batch}\n",
    "        session.run(optimizer, feed_dict=feed_dict_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "feed_dict_test = {x: data.test.images,\n",
    "                 y_true: data.test.labels,\n",
    "                 y_true_cls: data.test.cls}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "def print_accuracy():\n",
    "    acc = session.run(accuracy, feed_dict=feed_dict_test)\n",
    "    print('Accuracy on test=set: {0:.1%}'.format(acc))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on test=set: 9.8%\n"
     ]
    }
   ],
   "source": [
    "print_accuracy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on test=set: 78.3%\n",
      "Wall time: 132 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "optimize(num_iterations=10)\n",
    "print_accuracy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on test=set: 91.9%\n",
      "Wall time: 583 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "optimize(num_iterations=990)\n",
    "print_accuracy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "def print_confusion_matrix():\n",
    "    \n",
    "    cls_pred = session.run(y_pred_cls, feed_dict=feed_dict_test)\n",
    "    \n",
    "    cm = confusion_matrix(y_pred=cls_pred, y_true=data.test.cls)\n",
    "    \n",
    "    print(cm)\n",
    "    \n",
    "    plt.imshow(cm, interpolation='nearest', cmap=plt.cm.BuPu)\n",
    "    \n",
    "    plt.tight_layout()\n",
    "    plt.colorbar()\n",
    "    tick_marks = np.arange(args.num_classes)\n",
    "    plt.xticks(tick_marks)\n",
    "    plt.yticks(tick_marks)\n",
    "    \n",
    "    plt.xlabel('Predicted')\n",
    "    plt.ylabel('True')\n",
    "    \n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 961    0    3    3    0    3    7    1    2    0]\n",
      " [   0 1107    3    2    0    3    4    1   15    0]\n",
      " [   8    4  933   13   12    1   10   12   31    8]\n",
      " [   3    0   28  921    0   19    2   10   17   10]\n",
      " [   1    2    7    2  916    1    9    2    8   34]\n",
      " [  10    3    9   46    9  758   14    7   28    8]\n",
      " [  12    3    4    2   11   15  909    1    1    0]\n",
      " [   3    9   26    8    9    0    0  939    2   32]\n",
      " [   8   10   10   29    9   26   11   13  845   13]\n",
      " [  12    7    3   13   36    7    0   20    7  904]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print_confusion_matrix()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_weights():\n",
    "    w = session.run(weights)\n",
    "    \n",
    "    w_min = np.min(w)\n",
    "    w_max = np.max(w)\n",
    "    \n",
    "    fig, axes = plt.subplots(2, 5)\n",
    "    #fig.subplots_adjust(hspace = 0, wspace = 0.1)\n",
    "    \n",
    "    for i, ax in enumerate(axes.flat):\n",
    "        image = w[:, i].reshape(args.img_shape)\n",
    "\n",
    "        ax.set_xlabel('Weights: {0}'.format(i))\n",
    "        ax.imshow(image, vmin = w_min, vmax = w_max, cmap = plt.cm.seismic)\n",
    "            \n",
    "        ax.set_xticks([])\n",
    "        ax.set_yticks([])\n",
    "    \n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 10 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_weights()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on test=set: 92.6%\n",
      "Wall time: 1min 5s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "optimize(num_iterations=100000)\n",
    "print_accuracy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 10 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_weights()"
   ]
  },
  {
   "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
}