[PYTHON/TENSORFLOW] 단일 계층 퍼셉트론 신경망 만들기 (MNIST)

■ 단일 계층 퍼셉트론 신경망을 만드는 방법을 보여준다.

▶ input_data.py

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import gzip
import os
import tempfile
import numpy
from six.moves import urllib
from six.moves import xrange
import tensorflow as tf
from tensorflow.contrib.learn.python.learn.datasets.mnist import read_data_sets

 

▶ slp.py

import input_data
import matplotlib.pyplot as pp
import tensorflow as tf

numberPixelDatasets = input_data.read_data_sets("numberPixelDatasets_data/", one_hot = True)

learningRate       = 0.01
trainingEpochCount = 25
batchSize          = 100
displayStep        = 1

xTensor = tf.placeholder("float", [None, 784]) # 28×28 = 784
yTensor = tf.placeholder("float", [None, 10 ]) # 0-9 숫자 인식

weightVariable = tf.Variable(tf.zeros([784, 10]))
biasVariable   = tf.Variable(tf.zeros([10]))

activationTensor   = tf.nn.softmax(tf.matmul(xTensor, weightVariable) + biasVariable)
crossEntropyTensor = yTensor * tf.log(activationTensor)
costTensor         = tf.reduce_mean(-tf.reduce_sum(crossEntropyTensor, reduction_indices = 1))

optimizerOperation = tf.train.GradientDescentOptimizer(learningRate).minimize(costTensor)

averageList = []
epochList   = []

initializerOperation = tf.global_variables_initializer()

with tf.Session() as sess:
    sess.run(initializerOperation)
    for epoch in range(trainingEpochCount):
        averageCost = 0.
        totalBatch = int(numberPixelDatasets.train.num_examples / batchSize)
        for i in range(totalBatch):
            xBatchNDArray, yBatchNDArray = numberPixelDatasets.train.next_batch(batchSize)
            sess.run(optimizerOperation, feed_dict = {xTensor : xBatchNDArray, yTensor : yBatchNDArray})
            averageCost += sess.run(costTensor, feed_dict = {xTensor : xBatchNDArray, yTensor : yBatchNDArray}) / totalBatch
        if epoch % displayStep == 0:
            print("회차 :", '%04d' % (epoch + 1), "비용 =", "{:.9f}".format(averageCost))
        averageList.append(averageCost)
        epochList.append(epoch + 1)
    print("트레이닝 단계 완료")

    correctPredictionTensor = tf.equal(tf.argmax(activationTensor, 1), tf.argmax(yTensor, 1))
    accuracyTebsor          = tf.reduce_mean(tf.cast(correctPredictionTensor, "float"))
    print("모델 정확도 :", accuracyTebsor.eval({xTensor : numberPixelDatasets.test.images, yTensor : numberPixelDatasets.test.labels}))

    pp.plot(epochList, averageList, "o", label = "Logistic Regression Training phase")
    pp.ylabel("cost")
    pp.xlabel("epoch")
    pp.legend()
    pp.show()

 

▶ 결과

회차 : 0001 비용 = 1.176287905
회차 : 0002 비용 = 0.662353746
회차 : 0003 비용 = 0.550700741
회차 : 0004 비용 = 0.496796931
회차 : 0005 비용 = 0.463800235
회차 : 0006 비용 = 0.440929263
회차 : 0007 비용 = 0.423949052
회차 : 0008 비용 = 0.410716742
회차 : 0009 비용 = 0.399927922
회차 : 0010 비용 = 0.390965419
회차 : 0011 비용 = 0.383320482
회차 : 0012 비용 = 0.376762401
회차 : 0013 비용 = 0.371033744
회차 : 0014 비용 = 0.365946764
회차 : 0015 비용 = 0.361372409
회차 : 0016 비용 = 0.357258832
회차 : 0017 비용 = 0.353575538
회차 : 0018 비용 = 0.350136466
회차 : 0019 비용 = 0.347030454
회차 : 0020 비용 = 0.344138179
회차 : 0021 비용 = 0.341470252
회차 : 0022 비용 = 0.339007374
회차 : 0023 비용 = 0.336622848
회차 : 0024 비용 = 0.334463308
회차 : 0025 비용 = 0.332481194
트레이닝 단계 완료
모델 정확도 : 0.914