def printMatrix():
name = input(
'Input matrix\'s name to be printed. (Read from ".mtrx" file)\n>>>')
raw = IO.RAWReader()
raw.open(p.DATA_PATH + name + '.mtrx')
matrix = IO.getAMatrix(raw)
IO.printprettyMatrix(matrix)
def recoverNeuralNetwork():
name = input('Input NeuralNetwork\'s name to be recovered.\n>>>')
nn = NeuralNetwork.DFF()
nn.loadfromFile(name + '_backup')
nn.Name = name
IO.setValuetoConfigfile('setting.json', 'latestNN', nn.Name)
return nn
def editBackproprogation(MyNeuralNetwork):
speed = float(input('Input speed value.\n>>>'))
# targeterr = input('Input target error value.\n>>>')
IO.setValuetoConfigfile(
p.SAVED_PATH + MyNeuralNetwork.Name + '_profile.json',
'LearningAlgorithm', 'BackPropagation')
IO.setValuetoConfigfile(
p.SAVED_PATH + MyNeuralNetwork.Name + '_profile.json', 'Speed', speed)
def feedNeuralNetwork(MyNeuralNetwork):
# string = input('input "row(number of data amount)", "column(number of input layer\'s neuron size)"\n')
# string = string + ' ' + input('And then input "elements" row after row.\n')
string= '1 '+str(MyNeuralNetwork.LayerNeuronsCount[0])+' '
# Single set of data 1 * input size
string = string+input('Input single data set.(split by space and press enter)\n')
rawreader = IO.RAWReader()
rawreader.openString(string)
M = IO.getAMatrix(rawreader)
IO.printprettyMatrix(MyNeuralNetwork.feed(M))
def createNeuralNetwork():
name = input('Input NeuralNetwork\'s name to be created.\n>>>')
layerscount = int(input('Input "'+name+'" NeuralNetwork\'s layers count.\n>>>'))
layerneuronscount = []
for layer in range(0, layerscount):
layerneuronscount.append(int(input('Input layer('+str(layer+1)+'\\'+str(layerscount)+')\'s neurons count.\n>>>')))
nn = NeuralNetwork.DFF(layerscount, layerneuronscount, name)
nn.savetoFile()
IO.setValuetoConfigfile('setting.json', 'latestNN', nn.Name)
return nn
def editAdam(MyNeuralNetwork):
speed = float(input('Input speed value('+str(p.PROFILE_DEFAULT['Speed'])+').\n>>>'))
beta1 = float(input('Input Beta1(decay rate)('+str(p.PROFILE_DEFAULT['Beta1'])+').\n>>>'))
beta2 = float(input('Input Beta2(decay rate)('+str(p.PROFILE_DEFAULT['Beta2'])+').\n>>>'))
# targeterr = input('Input target error value.\n>>>')
IO.setValuetoConfigfile(p.SAVED_PATH+MyNeuralNetwork.Name+'_profile.json', 'LearningAlgorithm', 'Adam')
IO.setValuetoConfigfile(p.SAVED_PATH+MyNeuralNetwork.Name+'_profile.json', 'Speed', speed)
IO.setValuetoConfigfile(p.SAVED_PATH+MyNeuralNetwork.Name+'_profile.json', 'Beta1', beta1)
IO.setValuetoConfigfile(p.SAVED_PATH+MyNeuralNetwork.Name+'_profile.json', 'Beta2', beta2)
IO.setValuetoConfigfile(p.SAVED_PATH+MyNeuralNetwork.Name+'_profile.json', 'Epsilon', p.PROFILE_DEFAULT['Epsilon'])
def feedNeuralNetworkbymtrx(MyNeuralNetwork):
name = input('Input .mtrx\'s file name.\n')
rawreader = IO.RAWReader()
rawreader.open(p.DATA_PATH + name + '.mtrx')
M = IO.getAMatrix(rawreader)
np.set_printoptions(threshold=np.nan)
np.set_printoptions(precision=3)
np.set_printoptions(suppress=False)
IO.printprettyMatrix(MyNeuralNetwork.feed(M))
pass
def editRMSprop(MyNeuralNetwork):
speed = float(input('Input speed value('+str(p.PROFILE_DEFAULT['Speed'])+').\n>>>'))
decayrate = float(input('Input decay rate('+str(p.PROFILE_DEFAULT['DecayRate'])+').\n>>>'))
# targeterr = input('Input target error value.\n>>>')
IO.setValuetoConfigfile(p.SAVED_PATH+MyNeuralNetwork.Name+'_profile.json', 'LearningAlgorithm', 'RMSprop')
IO.setValuetoConfigfile(p.SAVED_PATH+MyNeuralNetwork.Name+'_profile.json', 'Speed', speed)
IO.setValuetoConfigfile(p.SAVED_PATH+MyNeuralNetwork.Name+'_profile.json', 'DecayRate', decayrate)
IO.setValuetoConfigfile(p.SAVED_PATH+MyNeuralNetwork.Name+'_profile.json', 'Epsilon', p.PROFILE_DEFAULT['Epsilon'])
def savetoFile(self, Filename=''):
if Filename == '':
Filename = self.Name
MyRAWWriter = IO.RAWWriter()
MyRAWWriter.append(int(self.LayersCount))
for layer in range(0, self.LayersCount):
MyRAWWriter.append(int(self.LayerNeuronsCount[layer]))
# Save each layer's neurons count one by one
MyRAWWriter.newline()
for layer in range(0, self.LayersCount - 1):
IO.writeAMatrix(self.Weight[layer], MyRAWWriter)
# Save each layer's weight one by one
for layer in range(0, self.LayersCount - 1):
IO.writeAMatrix(self.Bias[layer], MyRAWWriter)
# Save each layer's bias one by one
MyRAWWriter.write(p.SAVED_PATH + Filename + '.node')
def loadLatestNeuralNetwork():
name = IO.getValuefromConfigfile('setting.json', 'latestNN')
nn = NeuralNetwork.DFF()
if name != None:
nn.loadfromFile(name)
else:
print('No latest NeuralNetwork')
return nn
def loadfromFile(self, Filename):
try:
MyRAWReader = IO.RAWReader()
MyRAWReader.open(p.SAVED_PATH + Filename + '.node')
self.Name = Filename
self.LayersCount = int(MyRAWReader.pop())
self.LayerNeuronsCount = []
self.Weight = []
self.Bias = []
# Get the LayersCount first and Initlalize LayerNeuronsCount, Weight and Bias
for layer in range(0, self.LayersCount):
self.LayerNeuronsCount.append(int(MyRAWReader.pop()))
# Get each layer's neurons count one by one
for layer in range(0, self.LayersCount - 1):
self.Weight.append(IO.getAMatrix(MyRAWReader))
# Get each layer's weight one by one
for layer in range(0, self.LayersCount - 1):
self.Bias.append(IO.getAMatrix(MyRAWReader))
# Get each layer's bias one by one
except:
print('warning: Loading ' + Filename + '.node error!')
def editNesterovMomentum(MyNeuralNetwork):
speed = float(input('Input speed value.\n>>>'))
momentumrate = float(input('Input momentum rate.\n>>>'))
# targeterr = input('Input target error value.\n>>>')
IO.setValuetoConfigfile(p.SAVED_PATH+MyNeuralNetwork.Name+'_profile.json', 'LearningAlgorithm', 'NesterovMomentum')
IO.setValuetoConfigfile(p.SAVED_PATH+MyNeuralNetwork.Name+'_profile.json', 'Speed', speed)
IO.setValuetoConfigfile(p.SAVED_PATH+MyNeuralNetwork.Name+'_profile.json', 'Momentum_Rate', momentumrate)
def editAdaGrad(MyNeuralNetwork):
speed = float(input('Input speed value.\n>>>'))
# targeterr = input('Input target error value.\n>>>')
IO.setValuetoConfigfile(
p.SAVED_PATH + MyNeuralNetwork.Name + '_profile.json',
'LearningAlgorithm', 'AdaGrad')
IO.setValuetoConfigfile(
p.SAVED_PATH + MyNeuralNetwork.Name + '_profile.json', 'Speed', speed)
IO.setValuetoConfigfile(
p.SAVED_PATH + MyNeuralNetwork.Name + '_profile.json', 'Epsilon',
p.PROFILE_DEFAULT['Epsilon'])
def trainNeuralNetwork(MyNeuralNetwork):
error = float(input('Input target error(0.1)\n>>>'))
epochs = int(input('Input max epochs(-1)\n>>>'))
print('Getting Datas...')
Datas = IO.getDatas()
print('Datas getted')
if IO.getProfile(MyNeuralNetwork) == None:
print('You must have training profile first!')
editNeuralNetworkProfile(MyNeuralNetwork)
Profile = IO.getProfile(MyNeuralNetwork)
# Get Input/OutputData to matrix
if IO.getValuefromConfigfile('setting.json', 'Verbose') != None:
verbose = int(IO.getValuefromConfigfile('setting.json', 'Verbose'))
else:
verbose = p.VERBOSE_DEFAULT
if IO.getValuefromConfigfile('setting.json', 'Loop_per_N_times') != None:
loop = int(IO.getValuefromConfigfile('setting.json', 'Loop_per_N_times'))
else:
loop = p.VERBOSE_PER_LOOP_DEFAULT
# Setting from config file
TrainingType.trainbyBatch(MyNeuralNetwork, Datas, error, epochs, Profile, verbose, loop)
def setValuetoConfigfile():
listConfigfileValues()
name = input('Name Code:\n>>>')
value = input('Value:\n>>>')
IO.setValuetoConfigfile('setting.json', ConfigDict[str(name)], value)
def idx2mtrx():
FilenameIn = input('Input input IDX file name\n')
FilenameOut = input('Input output mtrx filename.\n')
IO.idx2mtrx(FilenameIn, FilenameOut)
def listConfigfileValues():
print('Config list:')
print('[v] Verbose Level. -> ' +
str(IO.getValuefromConfigfile('setting.json', 'Verbose')))
print('[n] Verbose per "N" times. -> ' +
str(IO.getValuefromConfigfile('setting.json', 'Loop_per_N_times')))
def getLatestNeuralNetworkName():
return IO.getValuefromConfigfile('setting.json', 'latestNN')
def loadNeuralNetwork():
name = input('Input NeuralNetwork\'s name to be loaded.\n>>>')
nn = NeuralNetwork.DFF()
nn.loadfromFile(name)
IO.setValuetoConfigfile('setting.json', 'latestNN', nn.Name)
return nn
def editTest1(MyNeuralNetwork):
# targeterr = input('Input target error value.\n>>>')
IO.setValuetoConfigfile(
p.SAVED_PATH + MyNeuralNetwork.Name + '_profile.json',
'LearningAlgorithm', 'Test1')
def printNeuralNetworkProfile(MyNeuralNetwork):
print(MyNeuralNetwork.Name + ' neuralnet\'s training profile:')
print(IO.getProfile(MyNeuralNetwork))