DL2-tensorfow之LR简单实践

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

随机生成1000个点，围绕在y=0.2x+0.4的直线周围

num_points = 1000
vectors_set = []
for i in range(num_points):
    x1 = np.random.normal(0.0, 0.6)
    y1 = x1 * 0.2 + 0.4 + np.random.normal(0.0, 0.03)
    vectors_set.append([x1, y1])

#原始数据集
x_data = [v[0] for v in vectors_set]
y_data = [v[1] for v in vectors_set]

#可视化
plt.figure(figsize=(12,8))
plt.scatter(x_data, y_data, c='r')
plt.show()

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

WARNING:tensorflow:From C:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\framework\op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.
Instructions for updating:
Colocations handled automatically by placer.

# 以预估值y和实际值y_data之间的均方误差作为损失
with tf.name_scope('loss'):
    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) #设置学习率参数

# 训练的过程就是最小化这个误差值
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 ('init_params: W =', sess.run(W), 'b =', sess.run(b), 'loss =', sess.run(loss))

#执行20次训练，并输出每次的执行结果
for step in range(20):
    sess.run(train)
    print ('step', step+1, ': W =', sess.run(W), 'b =', sess.run(b), 'loss =', sess.run(loss))

init_params: W = [-0.90279675] b = [0.] loss = 0.616048
step 1 : W = [-0.4868428] b = [0.39218372] loss = 0.18000357
step 2 : W = [-0.22638762] b = [0.39443102] loss = 0.07002382
step 3 : W = [-0.06461462] b = [0.39583817] loss = 0.027595185
step 4 : W = [0.03586529] b = [0.39671218] loss = 0.011226809
step 5 : W = [0.09827504] b = [0.39725503] loss = 0.004912118
step 6 : W = [0.13703877] b = [0.39759222] loss = 0.0024759993
step 7 : W = [0.16111557] b = [0.39780164] loss = 0.0015361789
step 8 : W = [0.17607008] b = [0.39793172] loss = 0.0011736094
step 9 : W = [0.18535858] b = [0.39801252] loss = 0.0010337348
step 10 : W = [0.19112782] b = [0.3980627] loss = 0.0009797732
step 11 : W = [0.1947112] b = [0.39809385] loss = 0.0009589559
step 12 : W = [0.19693689] b = [0.39811322] loss = 0.00095092464
step 13 : W = [0.1983193] b = [0.39812523] loss = 0.00094782625
step 14 : W = [0.19917794] b = [0.3981327] loss = 0.0009466311
step 15 : W = [0.19971125] b = [0.39813736] loss = 0.00094616995
step 16 : W = [0.2000425] b = [0.39814022] loss = 0.00094599195
step 17 : W = [0.20024824] b = [0.39814204] loss = 0.0009459233
step 18 : W = [0.20037603] b = [0.39814314] loss = 0.00094589696
step 19 : W = [0.20045541] b = [0.39814383] loss = 0.00094588666
step 20 : W = [0.20050472] b = [0.39814425] loss = 0.0009458827

plt.figure(figsize=(12,8))
plt.scatter(x_data, y_data, c='r')
plt.scatter(x_data, x_data * sess.run(W) + sess.run(b), c='b')
plt.show()