tensorflow学习

1. 将版本一的适用于版本二

将

import tensorflow as tf

替换为

import tensorflow.compat.v1 as tf 
tf.disable_v2_behavior() # 该行也可以不写

这样可以解决大部分的版本问题。

也可以单个更改，例如将

input1 = tf.placeholder(tf.float32)

替换为

input1 = tf.compat.v1.placeholder(tf.float32)

这样也可以更改。

还有一部分是

tf.mul替换为tf.multiply
tf.sub替换为tf.subtract
tf.neg替换为tf.negative

在网上找到，总结了一下

2. 使用tensorflow进行线性拟合

先生成样本

import numpy as np
import tensorflow.compat.v1 as tf
import matplotlib.pyplot as plt

# 随机生成1000个点，围绕在y = 0.1x+0.3的直线周围
num_points = 1000
vectors_set = []
for i in range(num_points):
    x1 = np.random.normal(0.0, 0.55)
    y1 = x1 * 0.1 + 0.3 + np.random.normal(0.0, 0.03)
    vectors_set.append([x1, y1])
    
# 生成一些样本
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()

 生成如下：

# 生成1维的W矩阵，取值是[-1,1]之间的随机数
W = tf.Variable(tf.random.uniform([1], -1.0, 1.0), name = 'W')
# 生成1维的b矩阵，初始值为0
b = tf.Variable(tf.zeros([1]), name = 'b')
# 经过计算得出预估值y
y = W * x_data + b

# 以预估值y和实际值y_data之间的均方误差作为损失
loss = tf.reduce_mean(tf.square(y - y_data), name = 'loss')
# 采用梯度下降法来优化参数
optimizer = tf.train.GradientDescentOptimizer(0.5)
# 训练的过程就是最小化这个误差值
train = optimizer.minimize(loss)

sess = tf.Session()

init = tf.global_variables_initializer()
sess.run(init)

# 初始化的W和b的值
print("W = ", sess.run(W), "b = ", sess.run(b), "loss = ", sess.run(loss))
# 执行20次训练
for step in range(20):
    sess.run(train)
    # 输出训练好的W和b
    print("W = ", sess.run(W), "b = ", sess.run(b), "loss = ", sess.run(loss))
# writer = tf.train.SummaryWriter("./tmp", sess.graph)

运行如下：

W =  [-0.330297] b =  [0.] loss =  0.28105444
W =  [-0.03630343] b =  [0.5151662] loss =  0.023182759
W =  [-0.09828708] b =  [0.36815736] loss =  0.004285659
W =  [-0.06601996] b =  [0.39915174] loss =  0.0025912393
W =  [-0.06004817] b =  [0.38301685] loss =  0.0021751502
W =  [-0.04800649] b =  [0.3800307] loss =  0.0018887866
W =  [-0.03850085] b =  [0.37400937] loss =  0.0016448094
W =  [-0.02916498] b =  [0.36925617] loss =  0.001432684
W =  [-0.02057617] b =  [0.36458784] loss =  0.0012479401
W =  [-0.0125269] b =  [0.3602931] loss =  0.0010870206
W =  [-0.00502343] b =  [0.3562681] loss =  0.00094685145
W =  [0.00198197] b =  [0.35251606] loss =  0.00082475675
W =  [0.00851948] b =  [0.3490131] loss =  0.00071840617
W =  [0.01462111] b =  [0.34574404] loss =  0.0006257689
W =  [0.02031574] b =  [0.34269297] loss =  0.00054507726
W =  [0.02563056] b =  [0.33984545] loss =  0.00047479055
W =  [0.03059087] b =  [0.3371878] loss =  0.00041356726
W =  [0.03522035] b =  [0.33470744] loss =  0.00036023854
W =  [0.03954105] b =  [0.3323925] loss =  0.0003137864
W =  [0.04357356] b =  [0.330232] loss =  0.00027332432
W =  [0.04733711] b =  [0.32821557] loss =  0.00023807972

可以看到结果是符合初始设置的线性函数的。

最后在加上

plt.scatter(x_data, y_data, c = 'r')
plt.plot(x_data,sess.run(W) * x_data + sess.run(b))
plt.show()

生成图片如下：

最后附上一篇比较好的有关线性回归的博客：使用tensorflow进行线性拟合