def deepLearningQ_training(Q, deviceName, epoch, printed):
# Q Table = [[[s0], [q00, q01, ...]], [[s1], [q10, q11, ...]], ...]
# convert to input = converted version of [[s0], [s1], ...]
# output = original version of [[q00, q01, ...], [q10, q11, ...], ...]
# input array (need to convert original array [s0])
inputData = []
for i in range(len(Q)):
inputData.append(stateTo1dArray(Q[i][0]))
# output array (as original)
outputData = []
for i in range(len(Q)):
outputData.append(Q[i][1])
# save input and output array as file
if len(inputData) > 0:
RD.saveArray('Q_input.txt', inputData)
if len(outputData) > 0:
RD.saveArray('Q_output.txt', outputData)
# train using deep learning and save the model (testInputFile and testOutputFile is None)
# need: modelConfig.txt
# DON'T NEED TO APPLY SIGMOID to training output data, because DLmain.deeplearning applies it
try:
DLmain.deepLearning('Q_input.txt', 'Q_output.txt', None, None, None,
None, 0.0, None, 'modelConfig.txt', deviceName,
epoch, printed, 'deepQ_model')
except:
print('Q_input.txt or Q_output.txt does not exist.')
def train(num, deviceName, epoch, printed):
# meta info
TRI = 'input_' + str(num) + '.txt'
TRO = 'output_' + str(num) + '.txt'
TEI = None
TEO = None
TE_real = None
TE_report = None
VAL_rate = 0.05
VAL_report = 'report_val_' + str(num) + '.txt'
modelConfig = 'model_config_' + str(num) + '.txt'
# training and test
# 'config.txt' is used for configuration
DL.deepLearning(TRI, TRO, TEI, TEO, TE_real, TE_report, VAL_rate,
VAL_report, modelConfig, deviceName, epoch, printed,
'model_' + str(num) + '_')
# main
if __name__ == '__main__':
# meta info
TRI = 'mnist_train_input.txt'
TRO = 'mnist_train_output.txt'
TEI = 'mnist_test_input.txt'
TEO = 'mnist_test_predict.txt'
TE_real = 'mnist_test_output.txt'
TE_report = 'mnist_test_report.txt'
VAL_rate = 0.0
VAL_report = 'mnist_val_report.txt'
modelConfig = 'mnist_model_config.txt'
# preprocess data
convertToNumeric()
# user data
deviceName = input('device name (for example, cpu:0 or gpu:0)')
epoch = int(input('epoch'))
printed = int(input('printed? (0 -> do not print)'))
# training and test
# 'config.txt' is used for configuration
DL.deepLearning(TRI, TRO, TEI, TEO, TE_real, TE_report, VAL_rate,
VAL_report, modelConfig, deviceName, epoch, printed,
'model')
int(math.sqrt(len(trainO_)))))
if validRate == 0.0:
print('\ntestI_array = stop - delta=' + str(i + 1) +
', index=' + str(j))
print(
testI_.reshape(int(math.sqrt(len(testI_))),
int(math.sqrt(len(testI_)))))
# deep learning : test array -> training array
#### validation mode ####
# test input is not used -> just use training data file to make model
# regardless of existance of test input file
if validRate > 0:
DL.deepLearning(trainIName, trainOName, None, None, None,
testReport, validRate, validReport, modelConfig,
deviceName, epoch, verbose, modelName)
#### use_n_sub == False or use_n_sub_for_test == True ####
# training input, training output and test input are all exist -> just use this data file to make model
elif use_n_sub == False or use_n_sub_for_test == True:
DL.deepLearning(trainIName, trainOName, testIName, testOName, None,
testReport, validRate, validReport, modelConfig,
deviceName, epoch, verbose, modelName)
#### use_n_sub == True and use_n_sub_for_test == False ####
# training input, training ouput and model exist, but test input does not exist
# -> read and reshape test data first as read.makeData
else:
# read and reshape test data from each file
if __name__ == '__main__':
# training input and test input:
# NORMALIZED using avg and stddev of each training input column
# meta info
TRI = 'train_input.txt'
TRO = 'train_output.txt'
TEI = 'test_input.txt'
TEO = 'test_output.txt'
TE_real = None
TE_report = 'report_test.txt'
VAL_rate = 0.0
VAL_report = 'report_val.txt'
modelConfig = 'model_config.txt'
# user data
deviceName = input('device name (for example, cpu:0 or gpu:0)')
epoch = int(input('epoch'))
printed = int(input('printed? (0 -> do not print)'))
# training and test
# 'config.txt' is used for configuration
for i in range(64):
DL.deepLearning(TRI, TRO, TEI, TEO[:-4] + '_' + str(i) + '.txt',
TE_real, TE_report[:-4] + '_' + str(i) + '.txt',
VAL_rate, VAL_report[:-4] + '_' + str(i) + '.txt',
modelConfig, deviceName, epoch, printed,
'model_' + str(i))
normalize(testOutputs_1, trainO_1_mean, trainO_1_std, 6), 1,
testNormalizedOutputFileName_1) # normalized test output (move)
SF.saveAsFile(
normalize(testOutputs_2, trainO_2_mean, trainO_2_std, 6), 1,
testNormalizedOutputFileName_2) # normalized test output (total cost)
## 딥러닝 실시
## 관련 학습/테스트 입출력 데이터 파일이 이미 존재하면 딥러닝에 의한 예측값만 파일로 쓰기
deviceName = input('device name (for example, cpu:0 or gpu:0)')
epoch = int(input('epoch'))
printed = int(input('printed? (0 -> do not print)'))
# comparison
print('\n ### deep learning for COMPARISON ###')
DL.deepLearning(trainInputFileName, trainNormalizedOutputFileName_0,
testInputFileName, testNormalizedOutputFileName_0, size,
deviceName, epoch, printed, 'test0_comparison')
# move
print('\n\n\n ### deep learning for MOVE ###')
DL.deepLearning(trainInputFileName, trainNormalizedOutputFileName_1,
testInputFileName, testNormalizedOutputFileName_1, size,
deviceName, epoch, printed, 'test1_move')
# total cost
print('\n\n\n ### deep learning for TOTAL COST ###')
DL.deepLearning(trainInputFileName, trainNormalizedOutputFileName_2,
testInputFileName, testNormalizedOutputFileName_2, size,
deviceName, epoch, printed, 'test2_totalCost')
## 딥러닝 결과로부터 loss 계산 : mean absolute and mean square error
def AIbase_deeplearning():
# read configuration file
f = open('config.txt', 'r')
fl = f.readlines()
f.close()
for i in range(len(fl)):
fl[i] = fl[i].split('\n')[0]
# init values as None
trainInputFile = None
trainOutputFile = None
testInputFile = None
testOutputFile = None
testOutputReal = None
test_report = None
valid_rate = None
valid_report = None
modelConfig = None
normalizeName = None
# extract configuration
# trainInput : train input data file name
# trainOutput : train output data file name
# testInput : test input data file name
# testOutput : test output (prediction) data file name
# testOutputReal : test output (real) data file name (if not None, compare output prediction with real test output data)
# test_report : test report file name
# valid_rate : rate of validation data among training data (0 -> use test data)
# valid_report : vaildation report file name
# modelConfig : model configuration file name
# normalizeName : normalize info (average and stddev) file name
for i in range(len(fl)):
configSplit = fl[i].split(' ') # split
if configSplit[0] == 'trainInput': trainInputFile = configSplit[1]
elif configSplit[0] == 'trainOutput': trainOutputFile = configSplit[1]
elif configSplit[0] == 'testInput': testInputFile = configSplit[1]
elif configSplit[0] == 'testOutput': testOutputFile = configSplit[1]
elif configSplit[0] == 'testOutputReal':
testOutputReal = configSplit[1]
elif configSplit[0] == 'test_report':
test_report = configSplit[1]
elif configSplit[0] == 'valid_rate':
valid_rate = float(configSplit[1])
elif configSplit[0] == 'valid_report':
valid_report = configSplit[1]
elif configSplit[0] == 'modelConfig':
modelConfig = configSplit[1]
elif configSplit[0] == 'normalizeName':
normalizeName = configSplit[1]
# convert 'None' to None
if trainInputFile == 'None': trainInputFile = None
if trainOutputFile == 'None': trainOutputFile = None
if testInputFile == 'None': testInputFile = None
if testOutputFile == 'None': testOutputFile = None
if testOutputReal == 'None': testOutputReal = None
if test_report == 'None': test_report = None
if valid_rate == 'None': valid_rate = None
if valid_report == 'None': valid_report = None
if modelConfig == 'None': modelConfig = None
if normalizeName == 'None': normalizeName = None
# do Deep Learning
# if the file for training/test data already exists, just write prediction for deep learning
deviceName = input('device name (for example, cpu:0 or gpu:0)')
epoch = int(input('epoch'))
printed = int(input('printed? (0 -> do not print)'))
print('\n ### deep learning ###')
DL.deepLearning(trainInputFile, trainOutputFile, testInputFile,
testOutputFile, testOutputReal, test_report, valid_rate,
valid_report, modelConfig, deviceName, epoch, printed,
'model')
