cnn_mnist.ipynb · Dingnan/卷积神经网络Mnist手写数字识别

加入 Gitee

与超过 1200万开发者一起发现、参与优秀开源项目，私有仓库也完全免费：）

该仓库未声明开源许可证文件（LICENSE），使用请关注具体项目描述及其代码上游依赖。

克隆/下载

cnn_mnist.ipynb 17.83 KB

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting /mnist/input_data\\train-images-idx3-ubyte.gz\n",
      "Extracting /mnist/input_data\\train-labels-idx1-ubyte.gz\n",
      "Extracting /mnist/input_data\\t10k-images-idx3-ubyte.gz\n",
      "Extracting /mnist/input_data\\t10k-labels-idx1-ubyte.gz\n",
      "step 100, entropy loss: 1.664484, l2_loss: 652.645386, total loss: 1.669053\n",
      "0.55\n",
      "step 200, entropy loss: 1.208336, l2_loss: 658.643982, total loss: 1.212946\n",
      "0.64\n",
      "step 300, entropy loss: 1.090859, l2_loss: 665.246155, total loss: 1.095516\n",
      "0.65\n",
      "step 400, entropy loss: 0.927444, l2_loss: 672.114380, total loss: 0.932148\n",
      "0.75\n",
      "step 500, entropy loss: 0.952603, l2_loss: 678.933899, total loss: 0.957356\n",
      "0.76\n",
      "step 600, entropy loss: 0.693806, l2_loss: 685.599609, total loss: 0.698605\n",
      "0.81\n",
      "step 700, entropy loss: 0.685850, l2_loss: 691.853394, total loss: 0.690693\n",
      "0.81\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-13-b113e1337c98>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m    117\u001b[0m     _, loss, l2_loss_value, total_loss_value = sess.run(\n\u001b[0;32m    118\u001b[0m                \u001b[1;33m[\u001b[0m\u001b[0mtrain_step\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcross_entropy\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0ml2_loss\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtotal_loss\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 119\u001b[1;33m                feed_dict={x: batch_xs, y_: batch_ys, learning_rate:lr, keep_prob:0.8})\n\u001b[0m\u001b[0;32m    120\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    121\u001b[0m     \u001b[1;32mif\u001b[0m \u001b[1;33m(\u001b[0m\u001b[0mstep\u001b[0m\u001b[1;33m+\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;33m%\u001b[0m \u001b[1;36m100\u001b[0m \u001b[1;33m==\u001b[0m \u001b[1;36m0\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Anaconda\\envs\\python3\\lib\\site-packages\\tensorflow\\python\\client\\session.py\u001b[0m in \u001b[0;36mrun\u001b[1;34m(self, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[0;32m    898\u001b[0m     \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    899\u001b[0m       result = self._run(None, fetches, feed_dict, options_ptr,\n\u001b[1;32m--> 900\u001b[1;33m                          run_metadata_ptr)\n\u001b[0m\u001b[0;32m    901\u001b[0m       \u001b[1;32mif\u001b[0m \u001b[0mrun_metadata\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    902\u001b[0m         \u001b[0mproto_data\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtf_session\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mTF_GetBuffer\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mrun_metadata_ptr\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Anaconda\\envs\\python3\\lib\\site-packages\\tensorflow\\python\\client\\session.py\u001b[0m in \u001b[0;36m_run\u001b[1;34m(self, handle, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[0;32m   1133\u001b[0m     \u001b[1;32mif\u001b[0m \u001b[0mfinal_fetches\u001b[0m \u001b[1;32mor\u001b[0m \u001b[0mfinal_targets\u001b[0m \u001b[1;32mor\u001b[0m \u001b[1;33m(\u001b[0m\u001b[0mhandle\u001b[0m \u001b[1;32mand\u001b[0m \u001b[0mfeed_dict_tensor\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1134\u001b[0m       results = self._do_run(handle, final_targets, final_fetches,\n\u001b[1;32m-> 1135\u001b[1;33m                              feed_dict_tensor, options, run_metadata)\n\u001b[0m\u001b[0;32m   1136\u001b[0m     \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1137\u001b[0m       \u001b[0mresults\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;33m[\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Anaconda\\envs\\python3\\lib\\site-packages\\tensorflow\\python\\client\\session.py\u001b[0m in \u001b[0;36m_do_run\u001b[1;34m(self, handle, target_list, fetch_list, feed_dict, options, run_metadata)\u001b[0m\n\u001b[0;32m   1314\u001b[0m     \u001b[1;32mif\u001b[0m \u001b[0mhandle\u001b[0m \u001b[1;32mis\u001b[0m \u001b[1;32mNone\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1315\u001b[0m       return self._do_call(_run_fn, feeds, fetches, targets, options,\n\u001b[1;32m-> 1316\u001b[1;33m                            run_metadata)\n\u001b[0m\u001b[0;32m   1317\u001b[0m     \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1318\u001b[0m       \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_do_call\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0m_prun_fn\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mhandle\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfeeds\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfetches\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Anaconda\\envs\\python3\\lib\\site-packages\\tensorflow\\python\\client\\session.py\u001b[0m in \u001b[0;36m_do_call\u001b[1;34m(self, fn, *args)\u001b[0m\n\u001b[0;32m   1320\u001b[0m   \u001b[1;32mdef\u001b[0m \u001b[0m_do_call\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfn\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m*\u001b[0m\u001b[0margs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1321\u001b[0m     \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1322\u001b[1;33m       \u001b[1;32mreturn\u001b[0m \u001b[0mfn\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0margs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1323\u001b[0m     \u001b[1;32mexcept\u001b[0m \u001b[0merrors\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mOpError\u001b[0m \u001b[1;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1324\u001b[0m       \u001b[0mmessage\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mcompat\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mas_text\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0me\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mmessage\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Anaconda\\envs\\python3\\lib\\site-packages\\tensorflow\\python\\client\\session.py\u001b[0m in \u001b[0;36m_run_fn\u001b[1;34m(feed_dict, fetch_list, target_list, options, run_metadata)\u001b[0m\n\u001b[0;32m   1305\u001b[0m       \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_extend_graph\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1306\u001b[0m       return self._call_tf_sessionrun(\n\u001b[1;32m-> 1307\u001b[1;33m           options, feed_dict, fetch_list, target_list, run_metadata)\n\u001b[0m\u001b[0;32m   1308\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1309\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0m_prun_fn\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mhandle\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfeed_dict\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfetch_list\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Anaconda\\envs\\python3\\lib\\site-packages\\tensorflow\\python\\client\\session.py\u001b[0m in \u001b[0;36m_call_tf_sessionrun\u001b[1;34m(self, options, feed_dict, fetch_list, target_list, run_metadata)\u001b[0m\n\u001b[0;32m   1407\u001b[0m       return tf_session.TF_SessionRun_wrapper(\n\u001b[0;32m   1408\u001b[0m           \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_session\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0moptions\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfeed_dict\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfetch_list\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtarget_list\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1409\u001b[1;33m           run_metadata)\n\u001b[0m\u001b[0;32m   1410\u001b[0m     \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1411\u001b[0m       \u001b[1;32mwith\u001b[0m \u001b[0merrors\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mraise_exception_on_not_ok_status\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;32mas\u001b[0m \u001b[0mstatus\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "from __future__ import absolute_import\n",
    "from __future__ import division\n",
    "from __future__ import print_function\n",
    "\n",
    "import argparse\n",
    "import sys\n",
    "\n",
    "from tensorflow.examples.tutorials.mnist import input_data\n",
    "\n",
    "import tensorflow as tf\n",
    "\n",
    "# Import data\n",
    "data_dir = 'tmp/tensorflow/mnist/input_data'\n",
    "mnist = input_data.read_data_sets(data_dir, one_hot=True)\n",
    "\n",
    "# Define loss and optimizer\n",
    "x = tf.placeholder(tf.float32, [None, 784])\n",
    "y_ = tf.placeholder(tf.float32, [None, 10])\n",
    "learning_rate = tf.placeholder(tf.float32)\n",
    "\n",
    "# 定义batch normalization的参数字典\n",
    "batch_norm_params1 = {  \n",
    "                        'is_training': True,\n",
    "                        # 是否是在训练模式，如果是在训练阶段，将会使用指数衰减函数（衰减系数为指定的decay），\n",
    "                        'zero_debias_moving_mean': True,\n",
    "                        # 如果为True，将会创建一个新的变量对 'moving_mean/biased' and 'moving_mean/local_step'，\n",
    "                        # 默认设置为False，将其设为True可以增加稳定性\n",
    "                        'decay': 0.995,            \n",
    "                        # 该参数能够衡量使用指数衰减函数更新均值方差时，更新的速度，取值通常在0.999-0.99-0.9之间，值\n",
    "                        # 越小，代表更新速度越快，而值太大的话，有可能会导致均值方差更新太慢，而最后变成一个常量1，而\n",
    "                        # 这个值会导致模型性能较低很多\n",
    "                        'updates_collections': None,\n",
    "                        # 该参数有一个默认值，ops.GraphKeys.UPDATE_OPS，当取默认值时，slim会在当前批训练完成后再更新均\n",
    "                        # 值和方差，这样会存在一个问题，就是当前批数据使用的均值和方差总是慢一拍，最后导致训练出来的模\n",
    "                        # 型性能较差。所以，一般需要将该值设为None，这样slim进行批处理时，会对均值和方差进行即时更新\n",
    "                    \n",
    "                    }\n",
    "\n",
    "with tf.name_scope('reshape'):\n",
    "    x_image = tf.reshape(x, [-1, 28, 28, 1])\n",
    "\n",
    "# First convolutional layer - maps one grayscale image to 32 feature maps.\n",
    "with tf.name_scope('conv1'):\n",
    "    h_conv1 = tf.contrib.slim.conv2d(x_image, 64, [6,6],padding='SAME',\n",
    "                                     weights_initializer=tf.contrib.slim.xavier_initializer(uniform=True),\n",
    "                                     normalizer_fn=tf.contrib.slim.batch_norm,                     \n",
    "                                     normalizer_params=batch_norm_params1,\n",
    "                                     activation_fn=tf.nn.relu)\n",
    "\n",
    "# Pooling layer - downsamples by 2X.\n",
    "with tf.name_scope('pool1'):\n",
    "    h_pool1 = tf.contrib.slim.max_pool2d(h_conv1, [2,2], stride=2, padding='VALID')\n",
    "\n",
    "# Second convolutional layer -- maps 32 feature maps to 64.\n",
    "with tf.name_scope('conv2'):\n",
    "    h_conv2 = tf.contrib.slim.conv2d(h_pool1, 128, [6,6],padding='SAME',\n",
    "                                     weights_initializer=tf.contrib.slim.xavier_initializer(uniform=True),\n",
    "                                     normalizer_fn=tf.contrib.slim.batch_norm,                     \n",
    "                                     normalizer_params=batch_norm_params1,\n",
    "                                     activation_fn=tf.nn.relu)\n",
    "\n",
    "# Second pooling layer.\n",
    "with tf.name_scope('pool2'):\n",
    "    h_pool2 = tf.contrib.slim.max_pool2d(h_conv2, [2,2],stride=[2, 2], padding='VALID')\n",
    "\n",
    "\n",
    "# Fully connected layer 1 -- after 2 round of downsampling, our 28x28 image\n",
    "# is down to 7x7x64 feature maps -- maps this to 1024 features.\n",
    "with tf.name_scope('fc1'):\n",
    "    h_pool2_flat = tf.contrib.slim.avg_pool2d(h_pool2, h_pool2.shape[1:3],stride=[1, 1], padding='VALID')\n",
    "    h_fc1 = tf.contrib.slim.conv2d(h_pool2_flat, 1024, [1,1],\n",
    "                                   weights_initializer=tf.contrib.slim.xavier_initializer(uniform=True),\n",
    "                                   activation_fn=tf.nn.relu)\n",
    "\n",
    "# Dropout - controls the complexity of the model, prevents co-adaptation of\n",
    "# features.\n",
    "with tf.name_scope('dropout'):\n",
    "    keep_prob = tf.placeholder(tf.float32)\n",
    "    h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)\n",
    "\n",
    "# Map the 1024 features to 10 classes, one for each digit\n",
    "with tf.name_scope('fc2'):\n",
    "    y = tf.squeeze(tf.contrib.slim.conv2d(h_fc1_drop, 10, [1,1], \n",
    "                                          weights_initializer=tf.contrib.slim.xavier_initializer(uniform=True),\n",
    "                                          activation_fn=None))\n",
    "\n",
    "\n",
    "# The raw formulation of cross-entropy,\n",
    "#\n",
    "#   tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(tf.nn.softmax(y)),\n",
    "#                                 reduction_indices=[1]))\n",
    "#\n",
    "# can be numerically unstable.\n",
    "#\n",
    "# So here we use tf.nn.softmax_cross_entropy_with_logits on the raw\n",
    "# outputs of 'y', and then average across the batch.\n",
    "cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y_, logits=y))\n",
    "\n",
    "l2_loss = tf.add_n( [tf.nn.l2_loss(w) for w in tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES)] )\n",
    "\n",
    "gamma=7e-5\n",
    "total_loss = cross_entropy + gamma*l2_loss\n",
    "\n",
    "train_step = tf.train.GradientDescentOptimizer(learning_rate).minimize(total_loss)\n",
    "\n",
    "sess = tf.Session()\n",
    "init_op = tf.global_variables_initializer()\n",
    "sess.run(init_op)\n",
    "\n",
    "# Train\n",
    "batch_size=100\n",
    "for step in range(1500):\n",
    "    batch_xs, batch_ys = mnist.train.next_batch(batch_size)\n",
    "    lr = 0.3\n",
    "    _, loss, l2_loss_value, total_loss_value = sess.run(\n",
    "               [train_step, cross_entropy, l2_loss, total_loss], \n",
    "               feed_dict={x: batch_xs, y_: batch_ys, learning_rate:lr, keep_prob:0.8})\n",
    "  \n",
    "    if (step+1) % 100 == 0:\n",
    "        print('step %d, entropy loss: %f, l2_loss: %f, total loss: %f' % \n",
    "                (step+1, loss, l2_loss_value, total_loss_value))\n",
    "        \n",
    "        # Test trained model\n",
    "        correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))\n",
    "        accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))\n",
    "        \n",
    "        print(sess.run(accuracy, feed_dict={x: batch_xs, y_: batch_ys, keep_prob:0.8}))\n",
    "        \n",
    "    if (step+1) % 500 == 0:\n",
    "        print(\"accuracy on test is\",sess.run(accuracy, feed_dict={x: mnist.test.images,\n",
    "                                    y_: mnist.test.labels, keep_prob:0.8}))\n",
    "        \n",
    "for step in range(1500,3000):\n",
    "    batch_xs, batch_ys = mnist.train.next_batch(batch_size)\n",
    "    lr = 0.08\n",
    "    _, loss, l2_loss_value, total_loss_value = sess.run(\n",
    "               [train_step, cross_entropy, l2_loss, total_loss], \n",
    "               feed_dict={x: batch_xs, y_: batch_ys, learning_rate:lr, keep_prob:0.8})\n",
    "  \n",
    "    if (step+1) % 100 == 0:\n",
    "        print('step %d, entropy loss: %f, l2_loss: %f, total loss: %f' % \n",
    "                (step+1, loss, l2_loss_value, total_loss_value))\n",
    "        \n",
    "        # Test trained model\n",
    "        correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))\n",
    "        accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))\n",
    "        \n",
    "        print(sess.run(accuracy, feed_dict={x: batch_xs, y_: batch_ys, keep_prob:0.8}))\n",
    "        \n",
    "    if (step+1) % 500 == 0:\n",
    "        print(\"accuracy on test is\",sess.run(accuracy, feed_dict={x: mnist.test.images,\n",
    "                                    y_: mnist.test.labels, keep_prob:0.8}))"
   ]
  },
  {
   "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}

一键复制编辑原始数据按行查看历史

提交于 2018-11-15 15:26 . Upload cnn_mnist.ipynb

Python

https://gitee.com/Xiaoyaoyaoyaojing/seventh_weeks_homework.git

git@gitee.com:Xiaoyaoyaoyaojing/seventh_weeks_homework.git

Xiaoyaoyaoyaojing

seventh_weeks_homework

卷积神经网络Mnist手写数字识别

master

Dingnan / 卷积神经网络Mnist手写数字识别

简介

发行版

贡献者

近期动态

Dingnan / 卷积神经网络Mnist手写数字识别 .gitee-modal { width: 500px !important; }

简介

发行版

贡献者

近期动态

搜索帮助

Dingnan / 卷积神经网络Mnist手写数字识别