[PYTHON/TENSORFLOW] 다층 퍼셉트론 신경망 만들기 (MNIST) : 1계층 Softmax

■ 순방향 신경망(Feed-Forward Neural Network, FFNN)을 만드는 방법을 보여준다.

▶ 예제 코드 (PY)


import numpy as np
import random
import tensorflow as tf
import tensorflow.examples.tutorials.mnist as mnist

inputLayerNodeCount  = 784
outputLayerNodeCount = 10

learningRate = 0.5
epochCount   = 10
batchSize    = 100

summaryLogDirectoryPath = "log_mnist_1_layer_softmax"

mnistDatasets = mnist.input_data.read_data_sets("data", one_hot = True)

inputLayerTensor          = tf.placeholder(tf.float32, [None, inputLayerNodeCount], name = "input")
hiddenLayerWeightVariable = tf.Variable(tf.zeros([inputLayerNodeCount, outputLayerNodeCount]))
hiddenLayerBiasVariable   = tf.Variable(tf.zeros([outputLayerNodeCount]))
outputLayerOutputTensor   = tf.nn.softmax(tf.matmul(inputLayerTensor, hiddenLayerWeightVariable) + hiddenLayerBiasVariable, name = "output")
correctOutputTensor       = tf.placeholder(tf.float32, [None, outputLayerNodeCount])
costTensor                = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels = correctOutputTensor, logits = outputLayerOutputTensor))
correctPredictionTensor   = tf.equal(tf.argmax(outputLayerOutputTensor, 1), tf.argmax(correctOutputTensor, 1))
accuracyTensor            = tf.reduce_mean(tf.cast(correctPredictionTensor, tf.float32))
optimizerOperation        = tf.train.GradientDescentOptimizer(learningRate).minimize(costTensor)

tf.summary.scalar("cost"    , costTensor    )
tf.summary.scalar("accuracy", accuracyTensor)

summaryTensor = tf.summary.merge_all()

with tf.Session() as session:
    session.run(tf.global_variables_initializer())
    fileWriter = tf.summary.FileWriter(summaryLogDirectoryPath, graph = tf.get_default_graph())
    batchCount = int(mnistDatasets.train.num_examples / batchSize)
    for epoch in range(epochCount):
        for batch in range(batchCount):
            batchInputNDArray, batchCorrectOutputNDArray = mnistDatasets.train.next_batch(batchSize)
            _, summaryBytes = session.run([optimizerOperation, summaryTensor], feed_dict = {inputLayerTensor : batchInputNDArray, correctOutputTensor : batchCorrectOutputNDArray})
            fileWriter.add_summary(summaryBytes, epoch * batchCount + batch)
        print("Epoch : ", epoch)
    print("정확도 : ", accuracyTensor.eval(feed_dict = {inputLayerTensor : mnistDatasets.test.images, correctOutputTensor : mnistDatasets.test.labels}))
    print("학습이 완료되었습니다.")

    testImageIndex   = random.randint(0, mnistDatasets.test.images.shape[0])
    testImageNDArray = mnistDatasets.test.images[testImageIndex]

    classificationNDArray = session.run(tf.argmax(outputLayerOutputTensor, 1), feed_dict = {inputLayerTensor : [testImageNDArray]})

    print("정답 : ", np.argmax(mnistDatasets.test.labels[testImageIndex]))
    print("판단 : ", classificationNDArray[0])

    saver = tf.train.Saver()

    saveFilePath = saver.save(session, "data/mnist_1_layer_softmax.ckpt")

    print("모델이 저장되었습니다 : %s" % saveFilePath)

import numpy as np

import random

import tensorflow as tf

import tensorflow.examples.tutorials.mnist as mnist

inputLayerNodeCount = 784

outputLayerNodeCount = 10

learningRate = 0.5

epochCount = 10

batchSize = 100

summaryLogDirectoryPath = "log_mnist_1_layer_softmax"

mnistDatasets = mnist.input_data.read_data_sets("data", one_hot = True)

inputLayerTensor = tf.placeholder(tf.float32, [None, inputLayerNodeCount], name = "input")

hiddenLayerWeightVariable = tf.Variable(tf.zeros([inputLayerNodeCount, outputLayerNodeCount]))

hiddenLayerBiasVariable = tf.Variable(tf.zeros([outputLayerNodeCount]))

outputLayerOutputTensor = tf.nn.softmax(tf.matmul(inputLayerTensor, hiddenLayerWeightVariable) + hiddenLayerBiasVariable, name = "output")

correctOutputTensor = tf.placeholder(tf.float32, [None, outputLayerNodeCount])

costTensor = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels = correctOutputTensor, logits = outputLayerOutputTensor))

correctPredictionTensor = tf.equal(tf.argmax(outputLayerOutputTensor, 1), tf.argmax(correctOutputTensor, 1))

accuracyTensor = tf.reduce_mean(tf.cast(correctPredictionTensor, tf.float32))

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

tf.summary.scalar("cost" , costTensor )

tf.summary.scalar("accuracy", accuracyTensor)

summaryTensor = tf.summary.merge_all()

with tf.Session() as session:

session.run(tf.global_variables_initializer())

fileWriter = tf.summary.FileWriter(summaryLogDirectoryPath, graph = tf.get_default_graph())

batchCount = int(mnistDatasets.train.num_examples / batchSize)

for epoch in range(epochCount):

for batch in range(batchCount):

batchInputNDArray, batchCorrectOutputNDArray = mnistDatasets.train.next_batch(batchSize)

_, summaryBytes = session.run([optimizerOperation, summaryTensor], feed_dict = {inputLayerTensor : batchInputNDArray, correctOutputTensor : batchCorrectOutputNDArray})

fileWriter.add_summary(summaryBytes, epoch * batchCount + batch)

print("Epoch : ", epoch)

print("정확도 : ", accuracyTensor.eval(feed_dict = {inputLayerTensor : mnistDatasets.test.images, correctOutputTensor : mnistDatasets.test.labels}))

print("학습이 완료되었습니다.")

testImageIndex = random.randint(0, mnistDatasets.test.images.shape[0])

testImageNDArray = mnistDatasets.test.images[testImageIndex]

classificationNDArray = session.run(tf.argmax(outputLayerOutputTensor, 1), feed_dict = {inputLayerTensor : [testImageNDArray]})