Linear Regression

import tensorflow as tf
from sklearn import metrics, model_selection
from sklearn.datasets import load_boston
import numpy as np

dataset = load_boston()
print(dataset.keys())

test_size = 0.2
seed = 7
x, y = dataset['data'], dataset['target']
print("Info: ")
print("Dim(x): ", x.shape, "\n", "Dim(y): ", y.shape)
print("Average data: ", np.average(x, axis=0))
print("Average target: ", np.average(y))
x_train, x_test, y_train, y_test = model_selection.train_test_split(x, y, test_size=test_size, random_state=seed)
y_train = np.array([y_train]).T
y_test = np.array([y_test]).T

# build model

input_x = tf.placeholder(shape=[None, 13], dtype=tf.float64, name="Input_X")
input_y = tf.placeholder(dtype=tf.float64, name="Input_Y")

weights = tf.Variable(tf.random_normal([13,1], dtype=tf.float64), name="Weight")
bias = tf.Variable(0, dtype=tf.float64, name="Bias")

predict = tf.add(tf.matmul(input_x, weights), bias, name="Model")
cost = tf.reduce_sum(tf.square(input_y-predict))
optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.01, name="GradientDescent").minimize(cost, name="Train_min")

sess = tf.Session()
init = tf.global_variables_initializer()

# Train model
sess.run(init)
for i in range(10):
    c, _ = sess.run([cost, optimizer], feed_dict={input_x: x_train, input_y: y_train})
    print("Cost of ", i, " is ", c)

writter = tf.summary.FileWriter(logdir="/tmp/linear/1", graph=sess.graph)

w, b, pred = sess.run([weights, bias, predict], feed_dict={input_x: x_test})
print("Weight: ", w, "\n Bias: ", b, "\nAverage pred: ", np.mean(pred, axis=0))