TensorFlow线性回归demo

生成随机点代码：

import numpy as np
import tensorflow 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):
    '''np.random.normal(loc=0.0, scale=1.0, size=None)
       作用：
           生成高斯分布的概率密度随机数
       loc：float
            此概率分布的均值（对应着整个分布的中心centre）
       scale：float
            此概率分布的标准差（对应于分布的宽度，scale越大越矮胖，scale越小，越瘦高）
       size：int or tuple of ints
            输出的shape，默认为None，只输出一个值    
    '''
    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]

# c='r'代表色彩或颜色序列
plt.scatter(x_data, y_data, c='r')
plt.show()

执行结果：

根据随机生成的点，反过来推线性回归方程：

# 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]))
# 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_daya之间的均方误差作为损失。计算损失值有很多方式，本程序采用均方误差的方式
# 均方误差介绍：https://blog.csdn.net/imjaron/article/details/79127674
# tf.math.square(x,name=None), 功能：计算元素x的平方
# tf.reduce_mean([x])，计算[x]的平均值。详情参考我的另一篇博文“TensorFlow tf.reduce_mean函数”
loss = tf.reduce_mean(tf.square(y - y_data))
# 采用梯度下降法来优化参数.0.5为学习率
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是多少
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))

训练结果：

W= [0.2508092] b= [0.] loss= 0.09659688
W= [0.21083724] b= [0.29800314] loss = 0.0046218005
W= [0.17724653] b= [0.2987708] loss = 0.0027052292
W= [0.15381296] b= [0.2994159] loss = 0.0017722554
W= [0.13746311] b= [0.2998659] loss = 0.0013180837
W= [0.12605566] b= [0.3001799] loss = 0.0010969935
W= [0.11809656] b= [0.30039898] loss = 0.0009893668
W= [0.11254342] b= [0.30055183] loss = 0.0009369743
W= [0.10866893] b= [0.30065846] loss = 0.0009114696
W= [0.10596567] b= [0.30073288] loss = 0.00089905394
W= [0.10407957] b= [0.3007848] loss = 0.0008930099
W= [0.10276362] b= [0.300821] loss = 0.00089006766
W= [0.10184547] b= [0.30084628] loss = 0.0008886355
W= [0.10120486] b= [0.30086392] loss = 0.00088793825
W= [0.10075791] b= [0.30087623] loss = 0.00088759896
W= [0.10044607] b= [0.3008848] loss = 0.0008874337
W= [0.10022849] b= [0.3008908] loss = 0.00088735315
W= [0.10007668] b= [0.30089498] loss = 0.0008873141
W= [0.09997077] b= [0.3008979] loss = 0.000887295
W= [0.09989686] b= [0.30089992] loss = 0.0008872856
W= [0.09984531] b= [0.30090135] loss = 0.0008872812

可视化展示：

plt.scatter(x_data, y_data, c='r')
# plt.plot(x,y,format_string,**kwargs)：x轴数据，y轴数据，format_string控制曲线的格式字串。format_string 由颜色字符，风格字符，和标记字符 
plt.plot(x_data, sess.run(W) * x_data + sess.run(b))
plt.show()

可视化展示结果：

plt.plot函数参考：https://blog.csdn.net/cjcrxzz/article/details/79627483