multicampus linear regression example

linear_regression.py

import tensorflow as tf
import matplotlib.pyplot as plt
%matplotlib inline

data = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
x = tf.placeholder(tf.int32)

a = tf.Variable(5)
b = tf.Variable(2)

y = a*x + b

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    result = sess.run(y, {x: data})
    print(result)

plt.plot(result)

samples = 1000
data = tf.constant([i for i in range(samples)], tf.float32)
label = 0.2 * data + 2.4 + tf.random_normal(shape=[samples,], stddev=70)
target = 0.2 * data + 2.4

with tf.Session() as sess:
    _data, _label, _target = sess.run([data, label, target])

plt.scatter(_data, _label, 1)
plt.scatter(_data, _target, 1)

x = tf.placeholder(tf.float32)
y_ = tf.placeholder(tf.float32)

weight = tf.Variable(tf.truncated_normal(shape=[1,]))
bias = tf.Variable(tf.zeros(1))
y = x * weight + bias

loss = tf.losses.mean_squared_error(y_, y)
train_op = tf.train.GradientDescentOptimizer(1e-6).minimize(loss)

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    
    losses = []
    for i in range(10):
        _, _loss = sess.run([train_op, loss], {x: _data, y_: _label})
        losses.append(_loss)
        print(_loss)
        
    pred = sess.run(y, {x:_data})

plt.plot(losses)

plt.scatter(_data, pred, 1)
plt.scatter(_data, _target, 1)