Linear Regression: code implementation

import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt

num_points=1000
vectors_set=[] 
for i in range(num_points):
    x1= np.random.normal(0.0, 0.05)#均值为0,标准差为0.05的随机值
    y1= x1*0.2+0.6+np.random.normal(0.0, 0.01) #加上小范围的数据浮点
    vectors_set.append([x1,y1])  #将x,y的值统一到向量中

#生成的样本点
x_data=[v[0] for v in vectors_set]
y_data=[v[1] for v in vectors_set]

plt.scatter(x_data,y_data,c='r')
plt.show()

Here Insert Picture Description

#初始化1维矩阵W,取值在[-1,1]之间
W = tf.Variable(tf.random_uniform([1],-1.0,1.0), name='W')

#初始化1维矩阵b,取值为0
b = tf.Variable(tf.zeros([1]), name='b')

#计算预测值y
y = W*x_data+b

#计算损失值(预测值与真实值间的均方差)
loss = tf.reduce_mean(tf.square(y-y_data),name='loss')

#采用梯度下降优化参数(W,b)
optimizer = tf.train.GradientDescentOptimizer(0.5)#学习率为0.5

#最小化损失值
train = optimizer.minimize(loss,name='loss')

init_op = tf.global_variables_initializer()
with tf.Session() as sess:
    sess.run(init_op)
    print("W = " ,sess.run(W),"b = " ,sess.run(b),"loss = ", sess.run(loss))
    #训练1000次训练
    for step in range(1000):  #迭代步数
      sess.run(train)
      print("W = " ,sess.run(W),"b = ", sess.run(b),"loss = " ,sess.run(loss))
    plt.scatter(x_data,y_data,c='r') #描绘样本点
    plt.scatter(x_data,sess.run(W)*x_data+sess.run(b))#描绘回归结果
    plt.show()

Here Insert Picture Description......
Here Insert Picture Description
declining in training loss, W and b gradually stabilize.

Guess you like

Origin blog.csdn.net/qq_43660987/article/details/92071837