MNIST Dataset

# Lab 7 Learning rate and Evaluation
import tensorflow as tf
import matplotlib.pyplot as plt
import random
 
tf.set_random_seed(777)  # for reproducibility
 
# 텐서플로우에서 데이터 받아오기
from tensorflow.examples.tutorials.mnist import input_data
 
# Check out https://www.tensorflow.org/get_started/mnist/beginners for
# more information about the mnist dataset
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
 
nb_classes = 10
 
# MNIST data image of shape 28 * 28 = 784
X = tf.placeholder(tf.float32, [None, 784])
# 0 - 9 digits recognition = 10 classes
Y = tf.placeholder(tf.float32, [None, nb_classes])
 
W = tf.Variable(tf.random_normal([784, nb_classes]))
b = tf.Variable(tf.random_normal([nb_classes]))
 
# Hypothesis (using softmax)
hypothesis = tf.nn.softmax(tf.matmul(X, W) + b)
 
cost = tf.reduce_mean(-tf.reduce_sum(Y * tf.log(hypothesis), axis=1))
train = tf.train.GradientDescentOptimizer(learning_rate=0.1).minimize(cost)
 
# Test model
is_correct = tf.equal(tf.argmax(hypothesis, 1), tf.argmax(Y, 1))
# Calculate accuracy
accuracy = tf.reduce_mean(tf.cast(is_correct, tf.float32))
 
# parameters
num_epochs = 15
batch_size = 100
num_iterations = int(mnist.train.num_examples / batch_size)
 
with tf.Session() as sess:
    # Initialize TensorFlow variables
    sess.run(tf.global_variables_initializer())
    # Training cycle
    for epoch in range(num_epochs):
        avg_cost = 0
 
        for i in range(num_iterations):
            batch_xs, batch_ys = mnist.train.next_batch(batch_size)
            _, cost_val = sess.run([train, cost], feed_dict={X: batch_xs, Y: batch_ys})
						# 각 epoch 마다 평균 비용을 구하려면, 2번째 값을 더한 후 2번째 iteration으로 나눠줘야 한다.
            avg_cost += cost_val / num_iterations
 
        print("Epoch: {:04d}, Cost: {:.9f}".format(epoch + 1, avg_cost))
 
    print("Learning finished")
 
    # Test the model using test sets
		# sess.run() 말고 다른 방법으로 노드를 실행하는 법
		# 노드에다가, .eval() 이라는 함수를 호출하면 된다.
    print(
        "Accuracy: ",
        accuracy.eval(
            session=sess, feed_dict={X: mnist.test.images, Y: mnist.test.labels}
        ),
    )
 
'''
Epoch: 0001, Cost: 2.826302672
Epoch: 0002, Cost: 1.061668952
Epoch: 0003, Cost: 0.838061315
Epoch: 0004, Cost: 0.733232745
Epoch: 0005, Cost: 0.669279885
Epoch: 0006, Cost: 0.624611836
Epoch: 0007, Cost: 0.591160344
Epoch: 0008, Cost: 0.563868987
Epoch: 0009, Cost: 0.541745171
Epoch: 0010, Cost: 0.522673578
Epoch: 0011, Cost: 0.506782325
Epoch: 0012, Cost: 0.492447643
Epoch: 0013, Cost: 0.479955837
Epoch: 0014, Cost: 0.468893674
Epoch: 0015, Cost: 0.458703488
Learning finished
Accuracy:  0.8951
'''

Training epoch/batch

epoch

전체 데이터 셋을 한번 훈련한 것

batch

전체 데이터셋을 메모리에 올려 한번에 훈련하기 어려우므로, 이 녀석을 잘라서 사용하는 단위

Example

전체 데이터가 1000 개이다. 1 epoch를 돌리기 위해서는 batch_size 500으로 2번의 반복이 필요하다.