一、卷積神經網絡CNN簡介

卷積神經網絡（ConvolutionalNeuralNetwork，CNN）最初是為解決圖像識別等問題設計的，CNN現在的應用已經不限于圖像和視頻，也可用于時間序列信號，比如音頻信號和文本數據等。CNN作為一個深度學習架構被提出的最初訴求是降低對圖像數據預處理的要求，避免復雜的特征工程。在卷積神經網絡中，第一個卷積層會直接接受圖像像素級的輸入，每一層卷積（濾波器）都會提取數據中最有效的特征，這種方法可以提取到圖像中最基礎的特征，而后再進行組合和抽象形成更高階的特征，因此CNN在理論上具有對圖像縮放、平移和旋轉的不變性。

卷積神經網絡CNN的要點就是局部連接（LocalConnection）、權值共享（WeightsSharing）和池化層（Pooling）中的降采樣（Down-Sampling）。其中，局部連接和權值共享降低了參數量，使訓練復雜度大大下降并減輕了過擬合。同時權值共享還賦予了卷積網絡對平移的容忍性，池化層降采樣則進一步降低了輸出參數量并賦予模型對輕度形變的容忍性，提高了模型的泛化能力。可以把卷積層卷積操作理解為用少量參數在圖像的多個位置上提取相似特征的過程。

更多請參見：深度學習之卷積神經網絡CNN

二、TensorFlow代碼實現

#!/usr/bin/env python2 # -*- coding: utf-8 -*- """ Created on Thu Mar 9 22:01:46 2017  @author: marsjhao """  import tensorflow as tf from tensorflow.examples.tutorials.mnist import input_data  mnist = input_data.read_data_sets("MNIST_data/", one_hot=True) sess = tf.InteractiveSession()  def weight_variable(shape):  initial = tf.truncated_normal(shape, stddev=0.1) #標準差為0.1的正態分布  return tf.Variable(initial)  def bias_variable(shape):  initial = tf.constant(0.1, shape=shape) #偏差初始化為0.1  return tf.Variable(initial)  def conv2d(x, W):  return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME')  def max_pool_2x2(x):  return tf.nn.max_pool(x, ksize=[1, 2, 2, 1],        strides=[1, 2, 2, 1], padding='SAME')  x = tf.placeholder(tf.float32, [None, 784]) y_ = tf.placeholder(tf.float32, [None, 10]) # -1代表先不考慮輸入的圖片例子多少這個維度，1是channel的數量 x_image = tf.reshape(x, [-1, 28, 28, 1]) keep_prob = tf.placeholder(tf.float32)  # 構建卷積層1 W_conv1 = weight_variable([5, 5, 1, 32]) # 卷積核5*5，1個channel，32個卷積核，形成32個featuremap b_conv1 = bias_variable([32]) # 32個featuremap的偏置 h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1) # 用relu非線性處理 h_pool1 = max_pool_2x2(h_conv1) # pooling池化  # 構建卷積層2 W_conv2 = weight_variable([5, 5, 32, 64]) # 注意這里channel值是32 b_conv2 = bias_variable([64]) h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2) h_pool2 = max_pool_2x2(h_conv2)  # 構建全連接層1 W_fc1 = weight_variable([7*7*64, 1024]) b_fc1 = bias_variable([1024]) h_pool3 = tf.reshape(h_pool2, [-1, 7*7*64]) h_fc1 = tf.nn.relu(tf.matmul(h_pool3, W_fc1) + b_fc1) h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)  # 構建全連接層2 W_fc2 = weight_variable([1024, 10]) b_fc2 = bias_variable([10]) y_conv = tf.nn.softmax(tf.matmul(h_fc1_drop, W_fc2) + b_fc2)  cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(y_conv),             reduction_indices=[1])) train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy) correct_prediction = tf.equal(tf.arg_max(y_conv, 1), tf.arg_max(y_, 1)) accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))  tf.global_variables_initializer().run()  for i in range(20001):  batch = mnist.train.next_batch(50)  if i % 100 == 0:   train_accuracy = accuracy.eval(feed_dict={x:batch[0], y_:batch[1],              keep_prob: 1.0})   print("step %d, training accuracy %g" %(i, train_accuracy))  train_step.run(feed_dict={x: batch[0], y_: batch[1], keep_prob:0.5}) print("test accuracy %g" %accuracy.eval(feed_dict={x: mnist.test.images,          y_: mnist.test.labels, keep_prob: 1.0}))