TensorFlow 卷积神经网络（CNN）实现手写数字识别（Mnist）

#导入模块
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()

#定义权重
def weights_variable(shape):
    initial=tf.truncated_normal(shape,stddev=0.1)#stddev代表的是随机噪声
    return tf.Variable(initial)
#定义偏量
def bias_variable(shape):
    initial=tf.constant(0.1,shape=shape)
    return tf.Variable(initial)

#定义卷积
def conv2d(x,W):#x代表的是输入，W代表的是权重
    return tf.nn.conv2d(x,W,strides=[1,1,1,1],padding="SAME")#如果W为[5,5,1,32]里面的参数，前面代表的是卷积核的尺寸，第三个参数是颜色，最后一个是卷积核个数，padding代表卷积核的输出和原图的大小一样,
                                                                #strides代表的是卷积模板移动的步长，都是1代表的不会遗漏的划过图片的每一个像素点
#定义池化，池化有两种方式，一种是sum另一种是max
def max_pool_2x2(x):#这个是TensorFlow里面2*2的最大池化
    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])#特征值
y_=tf.placeholder(tf.float32,[None,10])#真是的label
x_image=tf.reshape(x,[-1,28,28,1])#卷积神经网络会用到空间结构的信息，需要将1D的图片转换成2D的图片，将1*784转换成28*28，-1代表的是样本数量不固定，1代表的是颜色数量


#第一个卷积层
W_conv1 = weights_variable([5,5,1,32])
b_conv1 = bias_variable([32])
h_conv1 = tf.nn.relu(conv2d(x_image,W_conv1)+ b_conv1)
h_pool1 = max_pool_2x2(h_conv1)

#第二个卷积层
W_conv2 = weights_variable([5,5,32,64])
b_conv2 = bias_variable([64])
h_conv2 = tf.nn.relu(conv2d(h_pool1,W_conv2)+ b_conv2)
h_pool2 = max_pool_2x2(h_conv2)
#这个是将2D在转换为1D


W_fc1 = weights_variable([7*7*64,1024])
b_fc1 = bias_variable(([1024]))
h_pool2_flat=tf.reshape(h_pool2,[-1,7*7*64])
h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat,W_fc1)+b_fc1)
#这个是来钢制过拟合的dropout
keep_prob=tf.placeholder(tf.float32)
h_fc1_drop=tf.nn.dropout(h_fc1,keep_prob)


#链接Softmax进行概率的输出
W_fc2 = weights_variable([1024,10])
b_fc2 = bias_variable(([10]))
y_conv=tf.nn.softmax(tf.matmul(h_fc1_drop,W_fc2)+b_fc2)

#这个是正确率
cross_accuracy=tf.reduce_mean(-tf.reduce_sum(y_*tf.log(y_conv),reduction_indices=[1]))

train_step=tf.train.AdamOptimizer(1e-4).minimize(cross_accuracy)

correct_prediction=tf.equal(tf.argmax(y_conv,1),tf.argmax(y_,1))
accurcy=tf.reduce_mean(tf.cast(correct_prediction,tf.float32))


tf.global_variables_initializer().run()

for i in range(20000):
    bacth_size=mnist.train.next_batch(50)
    if i %100 ==0:
        train_accuracy=accurcy.eval(feed_dict={x:bacth_size[0],y_:bacth_size[1],keep_prob:1.0})
        print("step %d,  training accurcy %g"%(i,train_accuracy))
    train_step.run(feed_dict={x:bacth_size[0],y_:bacth_size[1],keep_prob:0.5})
print("test accuracy %g"% accurcy.eval(feed_dict={x:mnist.test.images,y_:mnist.train.labels,keep_prob:1.0}))