用tensorflow训练一个简单的线性函数

最近参加了几个面试，才发现现在大多数公司用的深度学习框架还是 tensorflow，之前学的MxNet，Gluon都没啥实际用途，不过要是学习的话可以学学，但是考虑找工作还是要好好学习tensorflow和pytorch！
今天学习了如何构建一个简单的神经网络去训练一个函数即 y=Wx+b

import tensorflow as tf
import numpy as np

# creat data
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.global_variables_initializer()

### create tensorflow structure start ###
sess = tf.Session()
sess.run(init)                 # Very important
for step in range(401):
    sess.run(train)
    if step % 20 == 0:
        print(step, sess.run(Weights), sess.run(biases))

输出：

0 [0.85651326] [-0.12958847]
20 [0.29266128] [0.20072113]
40 [0.1480372] [0.27524632]
60 [0.11197735] [0.29382804]
80 [0.10298639] [0.2984611]
100 [0.10074461] [0.2996163]
120 [0.10018566] [0.29990435]
140 [0.1000463] [0.29997614]
160 [0.10001156] [0.29999405]
180 [0.10000287] [0.29999852]
200 [0.1000007] [0.29999965]
220 [0.10000017] [0.29999992]
240 [0.10000011] [0.29999995]
260 [0.10000011] [0.29999995]
280 [0.10000011] [0.29999995]
300 [0.10000011] [0.29999995]
320 [0.10000011] [0.29999995]
340 [0.10000011] [0.29999995]
360 [0.10000011] [0.29999995]
380 [0.10000011] [0.29999995]
400 [0.10000011] [0.29999995]

结果分析：
一共跑了401步，但是只输出400/20=20组数据，从结果可以看出训练的W极限逼近于0.1，而训练出的b极限逼近于0.3！