MNIST - 일반 딥러닝 모델

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
 
mnist = input_data.read_data_sets("./mnist/data/", one_hot=True)
 
 
 
# 데이터 모음
input_data_train = mnist.train.images
output_data_train = mnist.train.labels
 
 
# 데이터 셋 입력 텐서
X = tf.placeholder(tf.float32, [None, 784])
Y = tf.placeholder(tf.float32, [None, 10])
 
 
# 변수 입력 텐서
import random
 
W1 = tf.Variable(tf.random_normal([784, 256], stddev=0.1))
b1 = tf.Variable(tf.random_normal([256], stddev=0.1))
 
W2 = tf.Variable(tf.random_normal([256, 64], stddev=0.1))
b2 = tf.Variable(tf.random_normal([64], stddev=0.1))
 
W3 = tf.Variable(tf.random_normal([64, 10], stddev=0.1))
b3 = tf.Variable(tf.random_normal([10], stddev=0.1))
 
 
 
 
out_1 = tf.matmul(X, W1)+b1
out_1 = tf.nn.relu(out_1)
 
out_2 = tf.matmul(out_1, W2)+b2
out_2 = tf.nn.relu(out_2)
 
out_3 = tf.matmul(out_2, W3)+b3
result = tf.nn.softmax(out_3)
 
 
 
 
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=result, labels= Y), axis=0)
optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.3).minimize(cost)
 
correct = tf.equal(tf.arg_max(result, 1), tf.arg_max(Y, 1))
accuracy = tf.reduce_mean(tf.cast(correct, tf.float32), axis=0)
 
 
 
 
tf.set_random_seed(100)
with tf.Session() as sess:
  init = tf.global_variables_initializer()
  sess.run(init)
 
  epoch = 10
  batch_size = 1000
  batch_num = int(mnist.train.num_examples/batch_size)
  
  # 훈련 과정
  for i in range(epoch):
    avg_cost = 0
    
    for j in range(batch_num):
      batch_x, batch_y = mnist.train.next_batch(batch_size)
      cost_rst, optimizer_rst = sess.run([cost, optimizer], feed_dict={X: batch_x, Y: batch_y})      
      avg_cost += cost_rst
      
    avg_cost= avg_cost/batch_num
    print("cost>> ", avg_cost)
  
  # 최종 정확도 검사 - test data 평가하기
  acc = accuracy.eval(feed_dict={X: mnist.test.images, Y: mnist.test.labels})
  print("최종 정확도 : ", acc*100,"%")
 
  # 라벨 실제로 예측하기
  idx = random.randint(0, mnist.test.num_examples-1)
  
  label_n = mnist.test.labels[idx:idx+1]
  label_n = tf.arg_max(label_n, 1)
  label_n = sess.run(label_n)
  
  img = mnist.test.images[idx:idx+1]
  predict = sess.run(result, feed_dict={X: img})
  predict = tf.arg_max(predict, 1)
  predict_n = sess.run(predict)
 
  print("Real Num : ", label_n)
  print("predicted Num : ", predict_n)
 
 

1. 데이터 셋 : MNIST 

2. 레이어 갯수 : 레이어 3개

3. cost function : 크로스 엔트로피

4. optimizer : Gradient Descent 

5. 전체적인 라이브러리 : 텐서플로우 하나