import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
def full_connection():
"""
用全连接来对数字进行识别
:return:
"""
# 1.准备数据
mnist=input_data.read_data_sets("./temp",one_hot=True)
x=tf.placeholder(dtype=tf.float32,shape=(None,784))
y_true=tf.placeholder(dtype=tf.float32,shape=(None,10))
# 2.构建模型
Weights=tf.Variable(initial_value=tf.random_normal(shape=(784,10)))
bias=tf.Variable(initial_value=tf.random_normal(shape=[10]))
y_predict=tf.matmul(x,Weights)+bias
# 3.构造损失函数
error=tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y_true,logits=y_predict))
# 4.优化损失函数
optimizer=tf.train.GradientDescentOptimizer(learning_rate=0.1).minimize(error)
# 5.准确率计算
equal_list=tf.equal(tf.arg_max(y_true,1),
tf.arg_max(y_predict,1))
accuracy=tf.reduce_mean(tf.cast(equal_list,tf.float32))
# 初始化变量
init=tf.global_variables_initializer()
# 开启回话
with tf.Session() as sess:
sess.run(init)
# 创建事件文件
# 拿真实值
image,lable=mnist.train.next_batch(100)
print("训练前，损失为%f"% sess.run(error,feed_dict={x:image,y_true:lable}))
for i in range(100):
a,loss,accuracy_value=sess.run([optimizer,error,accuracy],feed_dict={x:image,y_true:lable})
print("第%d次的训练，损失为%f，准确率为%f,"% (i+1,loss,accuracy_value))
full_connection()
