使用TensorFlow实现简单的线性回归模型

首先导入各种TensorFlow等工具及设置画图的大小及字体

import numpy as np
import tensorflow as tf
import matplotlib.pyplot as plt
plt.rcParams['figure.figsize'] = (10.0, 8.0)
plt.rcParams['font.sans-serif'] = 'NSimSun,Times New Roman'

生成用于进行线性回归的模型的数据

# 随机生成100个点，围绕在y=3x+5的直线周围
num_points = 200
vectors_set = []
for i in range(num_points):
    x1 = np.random.uniform(-10, 25)
    y1 = x1 * 3 + 5 + np.random.normal(0.0, 5)
    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()

生成的数据及画出的点像图如下：

设置模型的原始数据，编写现行回归的训练模型代码，并使用梯度下降算法进行训练

W = tf.Variable(tf.random_uniform([1], -1.0, 1.0), name='W')
b = tf.Variable(tf.zeros([1]), name='b')
y = W * x_data + b
loss = tf.reduce_mean(tf.square(y - y_data), name='loss')
optimizer = tf.train.GradientDescentOptimizer(0.005)
train = optimizer.minimize(loss, name='train')
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), "\tb =", sess.run(b), "\tloss =", sess.run(loss))
# 执行20次训练
for step in range(1000):
    sess.run(train)
    # 输出训练好的W和b
    if(step % 50 == 0):
        print ("W =", sess.run(W), "\tb =", sess.run(b), "\tloss =", sess.run(loss))
print ("最终的结果 W =", sess.run(W), "\tb =", sess.run(b), "\tloss =", sess.run(loss))

训练过程显示的参数结果如下：

W = [ 0.26262569]  b = [ 0.]  loss = 1366.89
W = [ 4.73117256]  b = [ 0.23248257]  loss = 386.389
W = [ 3.15473843]  b = [ 1.66317821]  loss = 32.05
W = [ 3.10857034]  b = [ 2.7120173]  loss = 27.6106
W = [ 3.07622719]  b = [ 3.44677091]  loss = 25.432
W = [ 3.05356979]  b = [ 3.96149445]  loss = 24.3628
W = [ 3.03769755]  b = [ 4.32207775]  loss = 23.8381
W = [ 3.02657843]  b = [ 4.57468128]  loss = 23.5806
W = [ 3.01878905]  b = [ 4.7516408]  loss = 23.4542
W = [ 3.01333213]  b = [ 4.87560892]  loss = 23.3922
W = [ 3.00950956]  b = [ 4.96245146]  loss = 23.3617
W = [ 3.00683141]  b = [ 5.02328825]  loss = 23.3468
W = [ 3.00495553]  b = [ 5.06590843]  loss = 23.3395
W = [ 3.00364113]  b = [ 5.09576511]  loss = 23.3359
W = [ 3.00272036]  b = [ 5.11667919]  loss = 23.3341
W = [ 3.00207567]  b = [ 5.13133097]  loss = 23.3332
W = [ 3.00162363]  b = [ 5.14159679]  loss = 23.3328
W = [ 3.00130725]  b = [ 5.14878702]  loss = 23.3326
W = [ 3.00108552]  b = [ 5.15382385]  loss = 23.3325
W = [ 3.00093007]  b = [ 5.15735531]  loss = 23.3324
W = [ 3.00082135]  b = [ 5.1598258]  loss = 23.3324
最终的结果 W = [ 3.00074625]  b = [ 5.16152811]  loss = 23.3324

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