def main():
cuda = gpuInit()
np.random.seed(1)
num_iterations = 1000
count = 10 # default
nSample = 0
layerStart = 0
layerLimit = 8
learning_rate = 0
simpleNN = 0
while (True):
userInput = input()
cds = Decision.userInput(userInput)
commands = cds.commands
# print(cds.noErr)
# print(commands)
if cds.noErr:
first = commands[0]
if first == 0: #createDataSet
count = 10
second = commands[1]
third = commands[2]
if second == 0: #train
trainSet = dataSet.createSimpleSet(1, third, count)
nSample = trainSet.X.shape[1] #batch크기 또는 train set 수
elif second == 1: #test
testSet = dataSet.createSimpleSet(1, third, count)
print(" | create datsSet complete ")
print(" | nSample = ", third)
elif first == 1: #autoBuild
if len(commands) == 1:
layerStart = 0
layerLimit = 8
elif len(commands) == 2:
layerStart = commands[1]
layerLimit = max(8, layerStart)
elif len(commands) == 3:
layerStart = commands[1]
layerLimit = commands[2]
if nSample > 0:
package = NeuralNetwork.autoBuilder(trainSet,
nSample,
layerStart=0,
layerLimit=8,
developmentMode=True)
# autoBuilder가 learning rate와 hiddenLayer의 수를 알아서 정해준다. 작업이 오래 걸릴 수 있음.
simpleNN = package[0]
learning_rate = package[1]
print("learning_rate", learning_rate)
else:
print(" error : please load trainSet first")
print(" or create dataSet | use command createDataSet")
elif first == 2: #model
if len(commands) > 1:
second = commands[1]
if second == 0: #train
if (learning_rate != 0) and (simpleNN != 0):
simpleNN.training_with_regularization(
trainSet, num_iterations, learning_rate)
train_acc = simpleNN.getAccuracy(trainSet)
print('train set Accuracy: %f' % train_acc + '%')
else:
print(" error : please build NN first")
print(" | use command autoBuild")
elif second == 1: #test
if (simpleNN != 0):
try:
train_acc = simpleNN.getAccuracy(trainSet)
test_acc = simpleNN.getAccuracy(testSet)
print('train set Accuracy: %f' % train_acc +
'%')
print('test set Accuracy: %f' % test_acc + '%')
except:
print(" error : please load test set")
else:
print(" error : please build NN first")
print(" | use command autoBuild")
elif first == 3: #loadDataSet
second = commands[1]
third = commands[2]
count = 100
if second == 0: #train
trainSet = dataSet.loadDataSet(count, third)
nSample = trainSet.X.shape[1] #batch크기 또는 train set 수
elif second == 1: #test
testSet = dataSet.loadDataSet(count, third)
print(" | load datsSet complete ")
print(" | nSample = ", nSample)
