TensorFlow MNIST CNN LeNet5模型

代码

import tensorflow as tffrom tensorflow.examples.tutorials.mnist import input_datamnist = input_data.read_data_sets('MNIST_data', one_hot=True)sess = tf.InteractiveSession()#训练数据x = tf.placeholder("float", shape=[None, 784])#训练标签数据y_ = tf.placeholder("float", shape=[None, 10])#把x更改为4维张量，第1维代表样本数量，第2维和第3维代表图像长宽， 第4维代表图像通道数, 1表示黑白x_image = tf.reshape(x, [-1,28,28,1])#第一层：卷积层conv1_weights = tf.get_variable("conv1_weights", [5, 5, 1, 32], initializer=tf.truncated_normal_initializer(stddev=0.1)) #过滤器大小为5*5, 当前层深度为1， 过滤器的深度为32conv1_biases = tf.get_variable("conv1_biases", [32], initializer=tf.constant_initializer(0.0))conv1 = tf.nn.conv2d(x_image, conv1_weights, strides=[1, 1, 1, 1], padding='SAME') #移动步长为1, 使用全0填充relu1 = tf.nn.relu( tf.nn.bias_add(conv1, conv1_biases) ) #激活函数Relu去线性化#第二层：最大池化层#池化层过滤器的大小为2*2, 移动步长为2，使用全0填充pool1 = tf.nn.max_pool(relu1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')#第三层：卷积层conv2_weights = tf.get_variable("conv2_weights", [5, 5, 32, 64], initializer=tf.truncated_normal_initializer(stddev=0.1)) #过滤器大小为5*5, 当前层深度为32， 过滤器的深度为64conv2_biases = tf.get_variable("conv2_biases", [64], initializer=tf.constant_initializer(0.0))conv2 = tf.nn.conv2d(pool1, conv2_weights, strides=[1, 1, 1, 1], padding='SAME') #移动步长为1, 使用全0填充relu2 = tf.nn.relu( tf.nn.bias_add(conv2, conv2_biases) )#第四层：最大池化层#池化层过滤器的大小为2*2, 移动步长为2，使用全0填充pool2 = tf.nn.max_pool(relu2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')#第五层：全连接层fc1_weights = tf.get_variable("fc1_weights", [7 * 7 * 64, 1024], initializer=tf.truncated_normal_initializer(stddev=0.1)) #7*7*64=3136把前一层的输出变成特征向量fc1_baises = tf.get_variable("fc1_baises", [1024], initializer=tf.constant_initializer(0.1))pool2_vector = tf.reshape(pool2, [-1, 7 * 7 * 64])fc1 = tf.nn.relu(tf.matmul(pool2_vector, fc1_weights) + fc1_baises)#为了减少过拟合，加入Dropout层keep_prob = tf.placeholder(tf.float32)fc1_dropout = tf.nn.dropout(fc1, keep_prob)#第六层：全连接层fc2_weights = tf.get_variable("fc2_weights", [1024, 10], initializer=tf.truncated_normal_initializer(stddev=0.1)) #神经元节点数1024, 分类节点10fc2_biases = tf.get_variable("fc2_biases", [10], initializer=tf.constant_initializer(0.1))fc2 = tf.matmul(fc1_dropout, fc2_weights) + fc2_biases#第七层：输出层# softmaxy_conv = tf.nn.softmax(fc2)#定义交叉熵损失函数cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(y_conv), reduction_indices=[1]))#选择优化器，并让优化器最小化损失函数/收敛, 反向传播train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)# tf.argmax()返回的是某一维度上其数据最大所在的索引值，在这里即代表预测值和真实值# 判断预测值y和真实值y_中最大数的索引是否一致，y的值为1-10概率correct_prediction = tf.equal(tf.argmax(y_conv,1), tf.argmax(y_,1))# 用平均值来统计测试准确率accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))#开始训练sess.run(tf.global_variables_initializer())for i in range(10000):    batch = mnist.train.next_batch(100)    if i%100 == 0:        train_accuracy = accuracy.eval(feed_dict={x:batch[0], y_: batch[1], keep_prob: 1.0}) #评估阶段不使用Dropout        print("step %d, training accuracy %g" % (i, train_accuracy))    train_step.run(feed_dict={x: batch[0], y_: batch[1], keep_prob: 0.5}) #训练阶段使用50%的Dropout#在测试数据上测试准确率print("test accuracy %g" % accuracy.eval(feed_dict={x: mnist.test.images, y_: mnist.test.labels, keep_prob: 1.0}))

最终准确率

比直接使用SoftMax的91%要好很多。