代码
代码如下,备注也算详细,就不多BB了
# -*- coding: utf-8 -*-
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
# 生成训练数据(x,y)
data_num= 100
value_set = []
for i in range(data_num):
x1 = np.random.normal(0.0, 0.6)
# 后面np.random.normal(0.0,0.03)加入扰动误差
y1 = x1 * 0.2 + 0.5 + np.random.normal(0.0, 0.03)
value_set.append([x1, y1])
x_data = [item[0] for item in value_set]
y_data = [item[1] for item in value_set]
# 定义计算图
W = tf.Variable(tf.random_uniform([1], -1.0, 1.0), name="W")
b = tf.Variable(tf.zeros([1]), name='b')
# 计算预测值
y = W * x_data + b
# 定义损失函数(方差)
loss = tf.reduce_mean(tf.square(y - y_data), name='loss')
# 优化器
optimizer = tf.train.GradientDescentOptimizer(0.5)
# 优化参数
train = optimizer.minimize(loss, name='train')
# 执行计算
sess = tf.Session()
init = tf.global_variables_initializer()
sess.run(init)
# 训练迭代20次
for step in range(20):
sess.run(train)
print('W=', sess.run(W), 'b=', sess.run(b), 'loss=', sess.run(loss))
# 画图
plt.scatter(x_data, y_data, c='r')
plt.plot(x_data, sess.run(W) * x_data + sess.run(b))
plt.show()
运行结果
想看结果,自己执行去!