TensorFlow深度学习作业一----多元回归

下表是某种商品的需求量（y，吨）、价格（x1，元/千克）和消费者收入（x2，元）观测值。

序号	X1	X2	y
1	5	1000	100
2	7	600	75
3	6	1200	80
4	6	500	70
5	8	30	50
6	7	400	65
7	5	1300	90
8	4	1100	100
9	3	1300	110
10	9	300	60

请用线性回归模型实现对如下数据的预测：

序号	X1	X2	y
1	8	900
2	7	1000
3	5	900
4	4	1500
5	8	130

要预测的为表2的y值

import tensorflow as tf
import numpy as np
import pylab as plt

x1 = np.float32([5,7,6,6,8,7,5,4,3,9])
x2 = np.float32([1000,600,1200,500,30,400,1300,1100,1300,300])
y1 = np.float32([100,75,80,70,50,65,90,100,110,60])
plt.scatter(x1,y1)
#tf.truncated_normal()产生正态分布的随机数
w1 = tf.Variable(tf.truncated_normal([]),name = 'w1')
w2 = tf.Variable(tf.truncated_normal([]),name = 'w2')
b = tf.Variable(tf.truncated_normal([]),name = 'b')

#两个矩阵中的元素对应相乘
price = tf.multiply(w1,x1) + tf.multiply(w2,x2) + b

loss = tf.reduce_mean(tf.square(price - y1))

optimizer = tf.train.AdamOptimizer(0.2)

train = optimizer.minimize(loss)

init = tf.global_variables_initializer()

sess = tf.Session()
sess.run(init)
for i in range(20000):
    sess.run(train)
    if i % 1000 == 0:
        print('loss:',sess.run(loss))
w1_v,w2_v,b_v = sess.run([w1,w2,b])
sess.close()
yu_x1 = np.float32([8,7,5,4,8])
yu_y1 = np.float32([900,1000,900,1500,130])
print("w1_v:",w1_v,"w2_v:",w2_v,"b_v:",b_v)
print("函数表达式为：y = ",w1_v,"×x1+",w2_v ,"×x2+",b_v)
for i in range(5):
    yu_y = yu_x1[i]*w1_v + yu_y1[i]*w2_v + b_v
    print(yu_y)
y = w1_v * x1 + w2_v + b_v
plt.plot(x1,y)
plt.show()

输出：

loss: 427720.06
loss: 138.73581
loss: 121.563866
loss: 97.33757
loss: 70.15316
loss: 47.760784
loss: 37.459503
loss: 33.808277
loss: 31.81664
loss: 31.200138
loss: 30.957703
loss: 30.863361
loss: 30.82673
loss: 30.822744
loss: 30.806988
loss: 30.804804
loss: 30.803982
loss: 30.805294
loss: 32.933388
loss: 30.803446
w1_v: -6.4932446 w2_v: 0.016316965 b_v: 106.33415
函数表达式为：y =  -6.4932446 ×x1+ 0.016316965 ×x2+ 106.33415
预测的y值为:
69.07346
77.1984
88.55319
104.836624
56.5094