基于CNN的MNIST手寫數字識别

CNN的具體理論知識可到百度或CSDN的其他部落格中查找相關内容，下面主要給出完整的代碼（代碼源于“莫煩python”視訊，https://morvanzhou.github.io/tutorials/machine-learning/tensorflow/5-06-save/）：

"""
使用TF的CNN網絡分類MNIST
CNN結構：input——Conv1+maxpool——>Conv2+maxpool——>full Connet1+dropout——>full Connect2——>output
卷積核大小：5*5
圖像大小：28*28*1
CNN使圖像變化過程：28*28*1——28*28*32——14*14*32——14*14*64——7*7*64——1024——10

"""
#-*-coding:utf-8-*-
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
# number 1 to 10 data
mnist = input_data.read_data_sets('MNIST_data', one_hot=True)

# 計算正确率
def compute_accuracy(v_xs, v_ys):
    global prediction
    y_pre = sess.run(prediction, feed_dict={xs: v_xs, keep_prob: })
    correct_prediction = tf.equal(tf.argmax(y_pre,), tf.argmax(v_ys,))
    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
    result = sess.run(accuracy, feed_dict={xs: v_xs, ys: v_ys, keep_prob: })
    return result

def weight_variable(shape):
    initial = tf.truncated_normal(shape,stddev=)
    return tf.Variable(initial)

def bias_variable(shape):
    initial = tf.constant(,shape=shape)
    return tf.Variable(initial)

def conv2d(x, W):
    return tf.nn.conv2d(x,W,strides=[,,,],padding='SAME')

def max_pool_2x2(x):
    return tf.nn.max_pool(x,ksize=[,,,],strides=[,,,],padding='SAME')

# define placeholder for inputs to network
xs = tf.placeholder(tf.float32, [None, ])  # 28x28
ys = tf.placeholder(tf.float32, [None, ])
keep_prob = tf.placeholder(tf.float32)
x_image = tf.reshape(xs,[-,,,])
## conv1 layer ##
W_conv1 = weight_variable([,,,]) #patch(kernel size) 5×5,in size 1,out size 32
b_conv1 = bias_variable([])
h_conv1 = tf.nn.relu(conv2d(x_image,W_conv1)+b_conv1)
h_pool1 = max_pool_2x2(h_conv1)

## conv2 layer ##
W_conv2 = weight_variable([,,,]) #patch(kernel size) 5×5,in size 32,out size 64
b_conv2 = bias_variable([])
h_conv2 = tf.nn.relu(conv2d(h_pool1,W_conv2)+b_conv2)
h_pool2 = max_pool_2x2(h_conv2)

## func1 layer ##
W_fc1 = weight_variable([**,])
b_fc1 = bias_variable([])
h_pool2_flat = tf.reshape(h_pool2,[-,**])
h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat,W_fc1)+b_fc1)
h_fc1_drop = tf.nn.dropout(h_fc1,keep_prob)

## func2 layer ##
W_fc2 = weight_variable([,])
b_fc2 = bias_variable([])
prediction = tf.nn.softmax(tf.matmul(h_fc1_drop,W_fc2)+b_fc2)


# the error between prediction and real data
cross_entropy = tf.reduce_mean(-tf.reduce_sum(ys * tf.log(prediction),
                                              reduction_indices=[]))       # loss
train_step = tf.train.AdamOptimizer().minimize(cross_entropy)

sess = tf.Session()

init = tf.global_variables_initializer()
sess.run(init)

for i in range():
    batch_xs, batch_ys = mnist.train.next_batch()
    sess.run(train_step, feed_dict={xs: batch_xs, ys: batch_ys, keep_prob: })
    if i %  == :
        print(compute_accuracy(
            mnist.test.images[:], mnist.test.labels[:]))