Tensorflow 搭建自己的神经网络(莫烦python 教程）

最新推荐文章于 2025-05-24 09:14:43 发布

原创最新推荐文章于 2025-05-24 09:14:43 发布 · 736 阅读

3 ·

CC 4.0 BY-SA版权

文章标签：

#Tensorflow语言学习

Tensorflow 专栏收录该内容

4 篇文章

订阅专栏

1、例子2

先构造结构，再初始化这个结构，让它们活动起来

import tensorflow as tf
import numpy as np

x_data=np.random.rand(100).astype(np.float32)
y_data=x_data*0.1+0.3

#create tensorflow structure start

Weights=tf.Variable(tf.random_uniform([1],-1.0,1.0))

biases=tf.Variable(tf.zeros([1]))


y=Weights*x_data+biases

loss=tf.reduce_mean(tf.square(y-y_data))

optimizer=tf.train.GradientDescentOptimizer(0.5)
train=optimizer.minimize(loss)

init=tf.initialize_all_variables()

sess=tf.Session()
sess.run(init)  #很重要

for step in range(201):
    sess.run(train)
    if step%20==0:
        print(step,sess.run(Weights),sess.run(biases))#每隔20次，输出权重W和B

输出的结果是：

0 [0.6654811] [0.00397022]
20 [0.24498683] [0.22517873]
40 [0.13706154] [0.2808742]
60 [0.10947368] [0.29511106]
80 [0.10242166] [0.2987503]
100 [0.10061904] [0.29968056]
120 [0.10015826] [0.29991835]
140 [0.10004047] [0.29997912]
160 [0.10001035] [0.29999468]
180 [0.10000264] [0.29999864]
200 [0.10000066] [0.29999965]

2、Session会话控制

session:执行命令，对话的控制，实现创造好的结构上的特定点的功能

每次run一下，才执行（tensorflow结构）

其中：sess.close()有和没有差不多，有的话，更为整洁和系统一点

import tensorflow as tf
matrix1=tf.constant([[3,3]])
matrix2=tf.constant([[2],[2]])
product=tf.matmul(matrix1,matrix2)  #matrix multiply  np.dot()

#method1
sess=tf.Session()
result=sess.run(product)
print(result)
sess.close()

#method2
with tf.Session() as sess:
    result2=sess.run(product)
    print(result2)###这个方法会自动close,即，sess.close()

运行结果

[[12]]
[[12]]

3、Variable变量

一定要定义成是一个变量，他才是一个变量

import tensorflow as tf

state = tf.Variable(0,name='counter')    # our first variable in the "global_variable" set
#print(state.name)
one=tf.constant(1)

new_value = tf.add(state, one)

update = tf.assign(state, new_value)
init=tf.global_variables_initializer()

with tf.Session() as sess:
    # once define variables, you have to initialize them by doing this
    sess.run(init)
    for i in range(3):
        sess.run(update)
        print(sess.run(state))
        #如果有定义，variable,一定要有init,再run

#如果有定义，variable,一定要有init,再run

4、placeholder传入值

variable 已经规定好初始自，placeholder是运行时才传入初始值，先占坑，然后每一次从外界传入一个值

import tensorflow as tf

x1 = tf.placeholder(dtype=tf.float32, shape=None)
y1 = tf.placeholder(dtype=tf.float32, shape=None)
z1 = x1 + y1

x2 = tf.placeholder(dtype=tf.float32, shape=[2, 1])
y2 = tf.placeholder(dtype=tf.float32, shape=[1, 2])
z2 = tf.matmul(x2, y2)

with tf.Session() as sess:
    # when only one operation to run
    z1_value = sess.run(z1, feed_dict={x1: 1, y1: 2})
    print(z1_value)
    # when run multiple operations
    z1_value, z2_value = sess.run(
        [z1, z2],       # run them together
        feed_dict={
            x1: 3, y1: 2,
            x2: [[2], [2]], y2: [[3, 3]]
        })
    print(z1_value)
    print(z2_value)

结果是：

3.0
5.0
[[6. 6.]
[6. 6.]]