深度学习实战(四) 在MNIST上训练一个CNN

Build your own CNN and try to achieve the highest possible accuracy on MNIST.

import tensorflow as tf
import numpy as np
from datetime import datetime
import os

def shuffle_batch(X, y, batch_size):
    rnd_idx = np.random.permutation(len(X))
    n_batches = len(X) // batch_size
    for batch_idx in np.array_split(rnd_idx, n_batches):
        X_batch, y_batch = X[batch_idx], y[batch_idx]
        yield X_batch, y_batch

height = 28
width = 28
channels = 1
n_inputs = height * width
n_outputs = 10

(X_train, y_train), (X_test, y_test) = tf.keras.datasets.mnist.load_data()
X_train = X_train.astype(np.float32).reshape(-1, 28*28) / 255.0
X_test = X_test.astype(np.float32).reshape(-1, 28*28) / 255.0
y_train = y_train.astype(np.int32)
y_test = y_test.astype(np.int32)
X_valid, X_train = X_train[:5000], X_train[5000:]
y_valid, y_train = y_train[:5000], y_train[5000:]

with tf.name_scope("inputs"):
    X = tf.placeholder(tf.float32, shape=[None, n_inputs], name="X")
    X_reshaped = tf.reshape(X, shape=[-1, height, width, channels])
    y = tf.placeholder(tf.int32, shape=[None], name="y")

#input:[batch, 28, 28, 1]   
#output:[batch, 28, 28, 32] 
conv1 = tf.layers.conv2d(X_reshaped, filters=32, kernel_size=3,
                         strides=1, padding="SAME",
                         activation=tf.nn.relu, name="conv1")
#input:[batch, 28, 28, 32]   
#output:[batch, 14, 14, 32] 
pool1 = tf.nn.max_pool(conv1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="SAME")
#input:[batch, 14, 14, 32]   
#output:[batch, 14, 14, 64]
conv2 = tf.layers.conv2d(pool1, filters=64, kernel_size=3,
                         strides=1, padding="SAME",
                         activation=tf.nn.relu, name="conv2")
#input:[batch, 14, 14, 64]   
#output:[batch, 7, 7, 64] 
pool2 = tf.nn.max_pool(conv2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="SAME")

with tf.name_scope("pool2"):
    pool2_flat = tf.reshape(pool2, shape=[-1, 64 * 7 * 7])

with tf.name_scope("fc1"):
    fc1 = tf.layers.dense(pool2_flat, 64, activation=tf.nn.relu, name="fc1")

with tf.name_scope("output"):
    logits = tf.layers.dense(fc1, n_outputs, name="output")
    Y_proba = tf.nn.softmax(logits, name="Y_proba")

with tf.name_scope("train"):
    xentropy = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=logits, labels=y)
    loss = tf.reduce_mean(xentropy)
    optimizer = tf.train.AdamOptimizer()
    training_op = optimizer.minimize(loss)

with tf.name_scope("eval"):
    correct = tf.nn.in_top_k(logits, y, 1)
    accuracy = tf.reduce_mean(tf.cast(correct, tf.float32))

with tf.name_scope("init_and_save"):
    init = tf.global_variables_initializer()
    saver = tf.train.Saver()
    
n_epochs = 15
batch_size = 100

with tf.Session() as sess:
    init.run()
    for epoch in range(n_epochs):
        for X_batch, y_batch in shuffle_batch(X_train, y_train, batch_size):
            sess.run(training_op, feed_dict={ X: X_batch, y: y_batch})
        acc_train = accuracy.eval(feed_dict={ X: X_batch, y: y_batch})
        acc_test = accuracy.eval(feed_dict={X: X_test, y: y_test})
        print(epoch, "Train accuracy:", acc_train, "Test accuracy:", acc_test)

        save_path = saver.save(sess, "./my_mnist_model")  

log info：

0 Train accuracy: 0.98 Test accuracy: 0.9762
1 Train accuracy: 0.99 Test accuracy: 0.985
2 Train accuracy: 0.98 Test accuracy: 0.9873
3 Train accuracy: 1.0 Test accuracy: 0.9889
4 Train accuracy: 1.0 Test accuracy: 0.9883
5 Train accuracy: 1.0 Test accuracy: 0.9896
6 Train accuracy: 1.0 Test accuracy: 0.9901
7 Train accuracy: 0.99 Test accuracy: 0.9902
8 Train accuracy: 1.0 Test accuracy: 0.9904
9 Train accuracy: 1.0 Test accuracy: 0.9911
10 Train accuracy: 1.0 Test accuracy: 0.9923
11 Train accuracy: 1.0 Test accuracy: 0.9904
12 Train accuracy: 1.0 Test accuracy: 0.991
13 Train accuracy: 1.0 Test accuracy: 0.989
14 Train accuracy: 1.0 Test accuracy: 0.9907