linear_regression.py

import os
import numpy as np
import tensorflow as tf
import tqdm
import matplotlib.pyplot as plt
%matplotlib inline
import seaborn

from tensorflow.python.framework import ops
ops.reset_default_graph()


sess =tf.InteractiveSession()
# создадим выборку
x = np.linspace(0,10, 1000)
y = np.sin(x) + np.random.normal(size=len(x))

plt.plot(x,y)
plt.show()

# и разобьем её на тренировочную и контрольную части
train_idxes = np.random.choice(list(range(len(x))), 3 * len(x)//4)
test_idxes = np.array(range(len(x)))
test_idxes = np.delete(test_idxes, train_idxes)

X_Train = x[train_idxes]
Y_Train = y[train_idxes]

X_Test = x[test_idxes]
Y_Test = y[test_idxes]

#Создадим граф
x_ = tf.placeholder(name="input", shape=[None, 1], dtype=tf.float32)
y_ = tf.placeholder(name= "output", shape=[None, 1], dtype=tf.float32)

model_output = tf.Variable(tf.random_normal([1]), name='bias') + tf.Variable(tf.random_normal([1]), name='k') * x_ # k*x+b

loss = tf.reduce_mean(tf.pow(y_ - model_output, 2)) # функция потерь
gd = tf.train.GradientDescentOptimizer(0.0001) #оптимизатор
train_step = gd.minimize(loss)
sess.run(tf.global_variables_initializer())
n_epochs = 100
train_errors = []
test_errors = []
for i in tqdm.tqdm(range(n_epochs)): # 100
    _, train_err = sess.run([train_step, loss ], feed_dict={x_:X_Train.reshape((len(X_Train),1)) , y_: Y_Train.reshape((len(Y_Train),1))})
    train_errors.append(train_err)
    test_errors.append(sess.run(loss, feed_dict={x_:X_Test.reshape((len(X_Test),1)) , y_: Y_Test.reshape((len(Y_Test),1))}))

plt.plot(list(range(n_epochs)), train_errors, label = 'train' )
plt.plot(list(range(n_epochs)), test_errors, label='test')
plt.legend()
plt.savefig('lin_reg.png')
print(train_errors[:10])
print(test_errors[:10])
plt.show()
plt.plot(x, y)
plt.plot(x,sess.run(model_output, feed_dict={x_:x.reshape((len(x),1))}))
plt.savefig("lr_forward_pass.png")