Tensorflow上使用简单神经网络.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "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": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'1.2.1'"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tf.__version__\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "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": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Size of:\n",
      "- Training-set:\t\t55000\n",
      "- Test-set:\t\t10000\n",
      "- Validation-set:\t5000\n"
     ]
    }
   ],
   "source": [
    "print(\"Size of:\")\n",
    "print(\"- Training-set:\\t\\t{}\".format(len(data.train.labels)))\n",
    "print(\"- Test-set:\\t\\t{}\".format(len(data.test.labels)))\n",
    "print(\"- Validation-set:\\t{}\".format(len(data.validation.labels)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "data.test.labels[0:5, :]\n",
    "data.test.cls = data.test.labels.argmax(axis=1)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# We know that MNIST images are 28 pixels in each dimension.\n",
    "img_size = 28\n",
    " # Images are stored in one-dimensional arrays of this length.\n",
    "img_size_flat = img_size * img_size\n",
    " # Tuple with height and width of images used to reshape arrays.\n",
    "img_shape = (img_size, img_size)\n",
    " # Number of classes, one class for each of 10 digits.\n",
    "num_classes = 10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "x = tf.placeholder(tf.float32, [None, img_size_flat])\n",
    "y_true = tf.placeholder(tf.float64, [None, num_classes])\n",
    "y_true_cls = tf.placeholder(tf.int64, [None])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "INPUT_NODE = 784\n",
    "OUTPUT_NODE = 10\n",
    "LAYER1_NODE = 500\n",
    "REGULARAZTION_RATE = 0.0001\n",
    "def get_weight_variable(shape, regularizer):\n",
    "    weights = tf.get_variable(\"weights\", shape,initializer=tf.truncated_normal_initializer(stddev=0.1)) #生成截断正态分布的随机数,标准差为0.1\n",
    "    if regularizer != None:\n",
    "        tf.add_to_collection('losses', regularizer(weights))\n",
    "    return weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#define the forward network\n",
    "def inference(input_tensor, regularizer):\n",
    "    with tf.variable_scope('layer1'):#声明第一层神经网络的变量并完成前向传播过程\n",
    "        weights = get_weight_variable([INPUT_NODE, LAYER1_NODE], regularizer)\n",
    "        biases = tf.get_variable(\"biases\", [LAYER1_NODE], initializer=tf.constant_initializer(0.0))\n",
    "        layer1 = tf.nn.relu(tf.matmul(input_tensor, weights) + biases)  #tf.nn.relu是作为激活函数\n",
    "\n",
    "    with tf.variable_scope('layer2'):#声明第二层神经网络的变量并完成前向传播过程\n",
    "        weights = get_weight_variable([LAYER1_NODE, OUTPUT_NODE], regularizer)\n",
    "        biases = tf.get_variable(\"biases\", [OUTPUT_NODE], initializer=tf.constant_initializer(0.0))\n",
    "        layer2 = tf.matmul(layer1, weights) + biases\n",
    "    return layer2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "regularizer = tf.contrib.layers.l2_regularizer(REGULARAZTION_RATE)  \n",
    "logits = inference(x, None)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y_pred = tf.nn.softmax(logits)\n",
    "y_pred_cls = tf.argmax(y_pred, dimension=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits=logits, labels=y_true)\n",
    "cost = tf.reduce_mean(cross_entropy)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.5).minimize(cost)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "correct_prediction = tf.equal(y_pred_cls, y_true_cls)\n",
    "accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "session = tf.Session()\n",
    "session.run(tf.global_variables_initializer())\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "batch_size = 100\n",
    "def optimize(num_iterations):\n",
    "    for i in range(num_iterations):\n",
    "        # Get a batch of training examples.\n",
    "        # x_batch now holds a batch of images and\n",
    "        # y_true_batch are the true labels for those images.\n",
    "        x_batch, y_true_batch = data.train.next_batch(batch_size)\n",
    "        # Put the batch into a dict with the proper names\n",
    "        # for placeholder variables in the TensorFlow graph.\n",
    "        # Note that the placeholder for y_true_cls is not set\n",
    "        # because it is not used during training.\n",
    "        feed_dict_train = {x: x_batch, y_true: y_true_batch}\n",
    "        # Run the optimizer using this batch of training data.\n",
    "        # TensorFlow assigns the variables in feed_dict_train\n",
    "        # to the placeholder variables and then runs the optimizer.\n",
    "        session.run(optimizer, feed_dict=feed_dict_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": true
   },
   "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": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "optimize(num_iterations=1)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def print_accuracy():\n",
    "    # Use TensorFlow to compute the accuracy.\n",
    "    acc = session.run(accuracy, feed_dict=feed_dict_test)\n",
    "    # Print the accuracy.\n",
    "    print(\"Accuracy on test-set: {0:.1%}\".format(acc))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on test-set: 19.8%\n"
     ]
    }
   ],
   "source": [
    "print_accuracy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on test-set: 26.8%\n"
     ]
    }
   ],
   "source": [
    "optimize(num_iterations=1)\n",
    "print_accuracy()\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on test-set: 95.2%\n"
     ]
    }
   ],
   "source": [
    "optimize(num_iterations=400)\n",
    "print_accuracy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on test-set: 97.0%\n"
     ]
    }
   ],
   "source": [
    "optimize(num_iterations=900)\n",
    "print_accuracy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "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": "3.5.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}