MNIST Dataset - CNN Coding - Part 2

Penyelesain MNIST data set dengan Convolutional Neural Network. Jika spec komputer tidak memadai, tidak disarankan menjalankan program ini, karena python akan crash.

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data

mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)


# helper funct init weigths
def init_weigths(shape):
    init_rand_dist = tf.truncated_normal(shape, stddev=0.1)
    return tf.Variable(init_rand_dist)


# helper funct init bias
def init_bias(shape):
    init_bias_val = tf.constant(0.1, shape=shape)
    return tf.Variable(init_bias_val)


# helper funct conv2d
def conv2d(x,W):
    #x -> [batch, h, w, channels]
    #w -> [filter h, filter w, channels in, channels out]
    return tf.nn.conv2d(x, W, strides=[1,1,1,1], padding='SAME')


# helper funct pooling
def max_pool_2by2(x):
    #x -> [batch, h, w, channels]
    #ksize ukuran window pooling 1,2,2,1 (angka 2 menunjukan h x w =  2 by 2), 
    #strides besar pergerakan (angka 2 menunjukan pergerakan 2 width dan 2 height)
    return tf.nn.max_pool(x, ksize=[1,2,2,1], strides=[1,2,2,1], padding='SAME')


#convolutional layer
def conv_layer(input_x, shape):
    W = init_weigths(shape)
    b = init_bias([shape[3]])
    return tf.nn.relu(conv2d(input_x, W) + b)


#normal layer
def norm_layer(input_layer, size):
    input_size = int(input_layer.get_shape()[1])
    W = init_weigths([input_size, size])
    b = init_bias([size])
    return tf.matmul(input_layer, W) + b


#placeholder
#shape: none-> berdasarkan jumlah data yang akan diinput, 784 adalah besar vector image (28x28)
x = tf.placeholder(tf.float32, shape=[None, 784])

y_true = tf.placeholder(tf.float32, shape=[None, 10])

#layers
#our input data, reshape to 28x28
x_image = tf.reshape(x, [-1, 28, 28, 1])

#ukuran 5x5, 1: grayscale, 32:feature
conv_1 = conv_layer(x_image, shape=[5,5,1,32])
conv_1_pool = max_pool_2by2(conv_1)

#32 diambil dari conv_1 32
conv_2 = conv_layer(conv_1_pool, shape=[5,5,32,64])
conv_2_pool = max_pool_2by2(conv_2)

conv_2_flat = tf.reshape(conv_2_pool, [-1, 7*7*64 ])
full_layer_one = tf.nn.relu(norm_layer(conv_2_flat, 1024))

#dropout
hold_prob = tf.placeholder(tf.float32)
full_one_dropout = tf.nn.dropout(full_layer_one, keep_prob=hold_prob)
y_pred = norm_layer(full_one_dropout, 10)

#optimze
cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y_true, logits=y_pred))

optimizer = tf.train.AdamOptimizer(learning_rate=0.01)
train = optimizer.minimize(cross_entropy)

init = tf.global_variables_initializer()

steps = 3000
with tf.Session() as sess:
    sess.run(init)
    for i in range(steps):
        batch_x, batch_y = mnist.train.next_batch(50)
        sess.run(train, feed_dict={x:batch_x, y_true:batch_y, hold_prob:0.5})
        
        if i%100 == 0:
            print("On step:{}", format(i))
            print("Accuracy: ")
            matches = tf.equal(tf.argmax(y_pred,1), tf.argmax(y_true,1))
            acc = tf.reduce_mean(tf.cast(matches, tf.float32))
            print(sess.run(acc, feed_dict={x:mnist.test.images, y_true:mnist.test.labels, hold_prob:1.0}))
            print('\n')