tensorflow使用多层感知器,要给神经网络加上隐含层,如下分别给出一层隐含层与两层隐含层的代码实现。
只有一层隐含层。
from tensorflow.examples.tutorials.mnist import input_data
import tensorflow as tf
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
sess = tf.InteractiveSession() #创建一个tensorflow默认的InteractiveSession,保证后面执行的时候就无需指定Session
in_units = 784
h1_units = 300
w1 = tf.Variable(tf.truncated_normal([in_units, h1_units], stddev=0.1)) #创建权重初始化为阶段的正态分布,其标准差为0.1
b1 = tf.Variable(tf.zeros([h1_units]))
w2 = tf.Variable(tf.zeros([h1_units, 10]))
b2 = tf.Variable(tf.zeros([10]))
x = tf.placeholder(tf.float32, [None, in_units])
keep_prob = tf.placeholder(tf.float32) #保留节点的概率
hidden1 = tf.nn.relu(tf.add(tf.matmul(x, w1), b1)) #实现一个激活函数为ReLU的隐含层
hidden1_drop = tf.nn.dropout(hidden1, keep_prob)
#调用tf.nn.dropout实现Dropout的功能,即随机将一部分的节点置为0,在一定程度上可以防止过拟合
y = tf.nn.softmax(tf.add(tf.matmul(hidden1_drop, w2), b2))
y_ = tf.placeholder(tf.float32, [None, 10])
cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(y), reduction_indices=[1]))
# global_step = tf.Variable(30, dtype=tf.int64) #定义全局步长,默认值为0
#train_step = tf.train.AdagradDAOptimizer(0.3, global_step=global_step).minimize(cross_entropy)
train_step = tf.train.AdagradOptimizer(0.3).minimize(cross_entropy)
tf.global_variables_initializer().run()
for i in range(3000):
batch_xs, batch_ys = mnist.train.next_batch(100)
train_step.run({x: batch_xs, y_: batch_ys, keep_prob: 0.75})
#测试
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
#tf.equal判断两个tensor是否每个元素都相等。返回一个格式为bool的tensor
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
#tf.cast将x的数据格式转化成dtype格式
print(accuracy.eval({x: mnist.test.images, y_: mnist.test.labels, keep_prob: 1.0}))
两层隐含层实现。
#两层隐藏层
from tensorflow.examples.tutorials.mnist import input_data
import tensorflow as tf
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
sess = tf.InteractiveSession() #创建一个tensorflow默认的InteractiveSession,保证后面执行的时候就无需指定Session
in_units = 784
h1_units = 300
h2_units = 300
w1 = tf.Variable(tf.truncated_normal([in_units, h1_units], stddev=0.1)) #创建权重初始化为阶段的正态分布,其标准差为0.1
b1 = tf.Variable(tf.zeros([h1_units]))
w2 = tf.Variable(tf.truncated_normal([h1_units, h2_units], stddev=0.1))
b2 = tf.Variable(tf.zeros([h2_units]))
#新增第二层隐藏层
w3 = tf.Variable(tf.zeros([h2_units, 10]))
b3 = tf.Variable(tf.zeros([10]))
x = tf.placeholder(tf.float32, [None, in_units])
keep_prob = tf.placeholder(tf.float32) #保留节点的概率
hidden1 = tf.nn.relu(tf.add(tf.matmul(x, w1), b1)) #实现一个激活函数为ReLU的隐含层
hidden1_drop = tf.nn.dropout(hidden1, keep_prob)
#调用tf.nn.dropout实现Dropout的功能,即随机将一部分的节点置为0,在一定程度上可以防止过拟合
#新增第二层隐藏层
hidden2 = tf.nn.relu(tf.add(tf.matmul(hidden1_drop, w2), b2)) #实现一个激活函数为ReLU的隐含层
hidden2_drop = tf.nn.dropout(hidden2, keep_prob)
y = tf.nn.softmax(tf.add(tf.matmul(hidden2_drop, w3), b3))
y_ = tf.placeholder(tf.float32, [None, 10])
cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(y), reduction_indices=[1]))
# global_step = tf.Variable(30, dtype=tf.int64) #定义全局步长,默认值为0
#train_step = tf.train.AdagradDAOptimizer(0.3, global_step=global_step).minimize(cross_entropy)
train_step = tf.train.AdagradOptimizer(0.3).minimize(cross_entropy)
tf.global_variables_initializer().run()
for i in range(3000):
batch_xs, batch_ys = mnist.train.next_batch(100)
train_step.run({x: batch_xs, y_: batch_ys, keep_prob: 0.75})
#测试
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
#tf.equal判断两个tensor是否每个元素都相等。返回一个格式为bool的tensor
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
#tf.cast将x的数据格式转化成dtype格式
print(accuracy.eval({x: mnist.test.images, y_: mnist.test.labels, keep_prob: 1.0}))