取自于TensorFlow实战
import tensorflow as tf from tensorflow.examples.tutorials.mnist import input_data #定义tensor的CPU运算优先级 import os os.environ['TF_CPP_MIN_LOG_LEVEL']='2' #导入书写数字输入级 mnist=input_data.read_data_sets('MNIST_data/',one_hot=True) #定义交互式会话框 sess=tf.InteractiveSession() #定义模型结构参数,输入、隐藏层节点数 in_units=784 h1_units=300 #定义权重、偏置 #w1截断正态分布 w1=tf.Variable(tf.truncated_normal([in_units,h1_units],stddev=0.1)) b1=tf.Variable(tf.zeros([h1_units])) w2=tf.Variable(tf.zeros([h1_units,10])) b2=tf.Variable(tf.zeros([10])) #定义输入、标签、dropout层保留率占位符 x=tf.placeholder(tf.float32,[None,784]) y_=tf.placeholder(tf.float32,[None,10]) keep_prob=tf.placeholder(tf.float32) #定义隐藏层、dropout层,输出层 hidden1=tf.nn.relu(tf.matmul(x,w1)+b1) hidden1_drop=tf.nn.dropout(hidden1,keep_prob) y=tf.nn.softmax(tf.matmul(hidden1_drop,w2)+b2) #定义损失函数为交叉熵 cross_entropy=tf.reduce_mean(-tf.reduce_sum(y_*tf.log(y),reduction_indices=[1])) #定义训练器 train=tf.train.AdagradOptimizer(0.3).minimize(cross_entropy) #初始化计算图 tf.global_variables_initializer().run() #定义训练过程,minibatch大小为100,dropout保持率为0.75 for i in range(3000): batch_x,batch_y=mnist.train.next_batch(100) train.run({x:batch_x,y_:batch_y,keep_prob:0.75}) #定义预测正误判断模型,(模型最大值index,标签最大值index)是否相等 correcct_prediction=tf.equal(tf.argmax(y,1),tf.argmax(y_,1)) #将预测正误判断模型bool型数据转换为float32,求平均为准确率 accuracy=tf.reduce_mean(tf.cast(correcct_prediction,tf.float32)) #使用eval计算训练好后的模型的正确率 print(accuracy.eval({x:mnist.test.images,y_:mnist.test.labels,keep_prob:1}))