"""
最終測試集結果0.9912
test accuracy 0.9912
"""
]import tensorflow as tf
from tensorflow.examples.tutorials.mnistimport input_data
def weight_variable(shape):? # 產(chǎn)生正態(tài)分布的卷積核權重，若生成的值與均值的差值大于兩倍的標準差，就重新生成
? ? initial= tf.truncated_normal(shape,stddev=0.1)
? ? return tf.Variable(initial)
def bias_variable(shape):? # 產(chǎn)生偏差矩陣
? ? initial= tf.constant(0.1,shape=shape)
? ? return tf.Variable(initial)
def conv2d(x, W):? # 卷積函數(shù)
? ? return tf.nn.conv2d(x, W,strides=[1,1,1,1],padding='SAME')? # 步長strides決定卷積的步伐，padding決定是否丟棄一部分，VALID丟棄，SAME不丟棄，不足之處補0
def max_pool_2x2(x):? # 池化函數(shù)
? ? return tf.nn.max_pool(x,ksize=[1,2,2,1],strides=[1,2,2,1],padding='SAME')
mnist= input_data.read_data_sets("MNIST_data/",one_hot=True)? # 下載或讀取數(shù)據(jù)集，one_hot編碼
x= tf.placeholder(tf.float32,[None,784])? # n行784列的輸入矩陣
W= tf.Variable(tf.zeros([784,10]))? # 權重矩陣，784行10列，初始化為零
b= tf.Variable(tf.zeros([10]))? # 偏置矩陣，10列的一個array
y_= tf.placeholder("float",[None,10])? # 訓練集的標簽
# 第一層卷積池化(采樣)得到14*14*32
W_conv1= weight_variable([5,5,1,32])
b_conv1= bias_variable([32])
x_image= tf.reshape(x,[-1,28,28,1])
h_conv1= tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)? # 首先是卷積加上偏置，卷積后再使用激活函數(shù)進行映射（實際是卷積層每個feature maps上的神經(jīng)元陣列進行卷積、映射）
h_pool1= max_pool_2x2(h_conv1)? # 實際上是進行采樣，減小數(shù)據(jù)量，提取主要特征
# 第二層卷積池化(采樣)得到7*7*64
W_conv2= weight_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)
# 全連接層將7*7*64展開
W_fc1= weight_variable([7 * 7 * 64,1024])? # 7*7*64個輸入，1024個神經(jīng)元
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)
# Dropout，減少過擬合
keep_prob= tf.placeholder('float')
h_fc1_drop= tf.nn.dropout(h_fc1, keep_prob)? # 一般用在全連接層,其含義是指，以keep_prob概率變?yōu)樵瓉淼?/keep_prob,以keep_prob概率變?yōu)?
# 輸出層,輸出到10個神經(jīng)元，構造softmax regression
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)
# 訓練和評估模型
cross_entropy= -tf.reduce_sum(y_*tf.log(y_conv))? # 計算交叉熵,用來衡量模型好壞
train_step= tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
correct_prediction= tf.equal(tf.argmax(y_conv,1), tf.argmax(y_,1))? # 正確的標簽與預測的標簽進行比對，確定正確率
accuracy= tf.reduce_mean(tf.cast(correct_prediction,"float"))? # argmax當axis=0時返回每一列的最大值的位置索引,當axis=1時返回每一行中的最大值的位置索引
sess= tf.InteractiveSession()? # 交互式環(huán)境比Session更加靈活
sess.run(tf.global_variables_initializer())
for iin range(20000):
? ? 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})
? ? ? ? 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}))
