ch9_mini_batch_gradient_descent.py

import tensorflow as tf
import numpy as np
from ch9_gradient_descent import load_data

def fetch_batch(X, y, batch_index, batch_size):
    i = batch_index * batch_size
    j = ((batch_index+1)* batch_size)-1
    X_batch = (X[i:j,])
    y_batch = (y[i:j]).reshape(-1, 1)
    return X_batch, y_batch


def main():

    print("Mini Batch Gradient Descent Example")
    X_train, y_train = load_data()
    m,n = X_train.data.shape

    batch_size = 5000
    n_batches = int(np.ceil(m / batch_size))
    n_epochs = 1000
    learning_rate = 0.01

    X = tf.placeholder(dtype=tf.float32, shape=(None, n), name="X")
    y = tf.placeholder(dtype=tf.float32, shape=(None, 1), name="y")

    theta = tf.Variable(tf.random_uniform([n, 1], -1.0, 1.0), name="theta")
    y_pred = tf.matmul(X, theta, name="predictions")

    error = y_pred - y
    mse = tf.reduce_mean(tf.square(error), name="mse")

    optimizer = tf.train.MomentumOptimizer(learning_rate=learning_rate, momentum=0.9)

    training_op = optimizer.minimize(mse)
    init = tf.global_variables_initializer()

    with tf.Session() as sess:
        sess.run(init)
        for epoch in range(n_epochs):
            for batch_index in range(n_batches):
                X_batch, y_batch = fetch_batch(X_train, y_train, batch_index, batch_size)
                sess.run(training_op, feed_dict={X: X_batch, y: y_batch})

            if epoch % 100 == 0:
                print("Epoch ", epoch, " MSE = ", mse.eval(feed_dict={X: X_batch, y: y_batch}))
        best_theta = theta.eval()
        print(best_theta)

if __name__ == "__main__":
    main()