python-tensorflow实现线性回归

import tensorflow as tf
import numpy as np

'''
tensorflow 实现线性回归
'''

#构造数据
w = tf.constant(3,tf.float32,name='w')
b = tf.constant(2,tf.float32,name='b')
x = tf.range(1.0,10.0,0.1,dtype=tf.float32,name='x')
y = w*x+b+tf.random_normal(shape=x.shape,mean=0.0,stddev=1.0,dtype=tf.float32,seed=0,name='y')

#构造模型
xx = x
ww = tf.get_variable(shape=[],dtype=tf.float32,name='ww',initializer=tf.constant_initializer(0.0))
bb = tf.get_variable(shape=[],dtype=tf.float32,name='bb',initializer=tf.constant_initializer(0.0))
yy = ww*xx+b

#构造损失函数 MSE
loss = tf.sqrt(tf.reduce_mean(tf.square(yy-y)))

#优化器
Optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.05).minimize(loss)   



#保存模型
saver = tf.train.Saver()
with tf.Session() as sess:
    tf.global_variables_initializer().run()

    for i in range(5):
        Optimizer.run()    #损失函数进行了一次求导+参数更新
        print("loss:%f,w:%f,b:%f"%(loss.eval(),w.eval(),b.eval()))

    saver.save(sess,'./model/linear_regression.ckpt')

#恢复模型
with tf.Session() as sess:
    tf.global_variables_initializer().run()
    saver.restore(sess,'./model/linear_regression.ckpt')

    print("w:%f,b:%f" % (w.eval(), b.eval()))

输出结果：

训练：

loss:16.177019,w:3.000000,b:2.000000
loss:14.493311,w:3.000000,b:2.000000
loss:12.697419,w:3.000000,b:2.000000
loss:10.922069,w:3.000000,b:2.000000
loss:8.975759,w:3.000000,b:2.000000

模型加载：

w:3.000000,b:2.000000