def train(self, dataset):
"""Train the network and record the accuracy.
Args:
dataset (str): Name of dataset to use.
"""
if self.accuracy == 0.:
self.accuracy = train_and_score(self.network, dataset)
def train(self, valid, i):
"""Train the network and record the accuracy.
"""
if self.accuracy == 0.:
self.avg_accuracy, self.accuracy = train_and_score(
self.network, self.B, valid, i)
def train(self,
config,
x_train=None,
y_train=None,
x_test=None,
y_test=None):
"""Train the network and record the accuracy.
Args:
config (str): instance of Config.
"""
self.model_type = config.model_type
if self.accuracy == 0.:
score, _ = train_and_score(config,
self,
x_train=x_train,
y_train=y_train,
x_test=x_test,
y_test=y_test)
self.accuracy = score[1]
print('\n\nACC: %.4f\n\n' % score[1])
if config.verbose == 1:
self.log(score)
elif config.verbose == 2 and config.min_acc <= score[1]:
self.log(score)
if config.min_acc <= score[1]:
network_to_json(self)
def train(self):
"""Train the network and record the mse.
Args:
dataset (str): Name of dataset to use.
"""
if self.score == 0.:
self.score = train_and_score(self.network)
def apply_crossval(x, y, Y, model, n_folds=10):
skf = StratifiedKFold(n_splits=n_folds, shuffle=True, random_state=1)
c = 0
log_dir = config.tb_log_dir
scores = []
for train, test in skf.split(x, Y):
print("Running Fold", c + 1, "/", n_folds)
if log_dir:
config.tb_log_dir = os.path.join(log_dir, 'fold%d' % (c))
keras_model = utils.json_to_model(model, config)
score, _ = train_and_score(config,
model=keras_model,
x_train=x[test],
y_train=y[test],
x_test=x[train],
y_test=y[train])
K.clear_session() # clear model
print('Acc: %.2f%% \nLoss: %.2f' % (score[1] * 100, score[0]))
scores.append(score[1] * 100)
logging.info('*' * 60)
logging.info(score)
logging.info('*' * 60)
c += 1
mean = np.mean(scores)
std = np.std(scores)
logging.info('Mean %f' % mean)
logging.info('Std %f' % std)
print('Acc: %.2f%% \t Std: %.2f' % (mean, std))
def train(self, dataset, nepochs):
"""Train the network and record the accuracy.`
"""
if self.accuracy == 0.:
acc, a_e = train_and_score(self.network, dataset, nepochs)
self.accuracy = acc
self.acc_epoch = a_e
def train(self):
"""
Train the network and record the accuracy.
:return:
"""
if self.accuracy == 0.:
self.accuracy = train_and_score(self.network)
def train(self, trainingset):
"""Train the genome and record the accuracy.
Args:
trainingset (str): Name of dataset to use.
"""
if self.accuracy == 0.0: #don't bother retraining ones we already trained
self.accuracy = train_and_score(self, trainingset)
def train(self, trainingset):
"""Train the genome and record the accuracy.
Args:
trainingset (str): Name of dataset to use.
"""
if self.r == -1.0:
self.r = train_and_score(self.geneparam, trainingset)
def train(self, dataset):
"""Train the network and record the accuracy.
Args:
dataset (str): Name of dataset to use.
"""
if self.accuracy == 0.:
self.accuracy = train_and_score(self.network, dataset)
def train(self, x_train, y_train, x_val, y_val, x_test, y_test,
batch_size):
"""Train the network and record the accuracy.
Args:
dataset (str): Name of dataset to use.
"""
if self.loss == 0.:
self.loss = train_and_score(self.network, x_train, y_train, x_val,
y_val, x_test, y_test, batch_size)
def train(config, path, x_train=None, y_train=None, x_test=None, y_test=None):
model = json_to_model(path, config)
score, h = train_and_score(config,
model=model,
x_train=x_train,
y_train=y_train,
x_test=x_test,
y_test=y_test)
return score, h
def train(self, dataset):
"""Realiza o treinamento da rede e retorna a acurácia da mesma.
Args:
dataset (str): path do dataset (conjunto de teste)
"""
print(dataset)
if self.accuracy == 0.:
self.accuracy = train_and_score(self.network, dataset)
def train(self, dataset, skip_real_train=False):
"""Train the network and record the accuracy.
Args:
dataset (str): Name of dataset to use.
"""
if self.accuracy == 0.:
if (skip_real_train == False):
self.accuracy = train_and_score(self.network, dataset)
else:
self.accuracy = random.randrange(0, 100) / 100
print("Test condition of each network : {0}".format(self.network))
print("Test result of each network : {0}".format(self.accuracy))
def train(self, dataset, count, total):
"""Train the network and record the accuracy.
Args:
dataset (str): Name of dataset to use.
"""
if self.accuracy == 0.:
print('Training Network (%2d/%2d) %s' % (count, total, self.print()))
self.loss, self.accuracy, self.model = train_and_score(self.network, dataset)
self.fit = 1 / self.loss if self.loss != 0 else sys.maxint
else:
print(' Using Network (%2d/%2d) %s - accuracy: %.2f%%, - loss: %.4f' % (count, total, self.print(), self.accuracy * 100, self.loss))
def train(self, dataset, count, total):
"""Train the network and record the accuracy.
Args:
dataset (str): Name of dataset to use.
"""
if self.accuracy == 0.:
print('Training Network (%2d/%2d) %s' %
(count, total, self.print()))
self.loss, self.accuracy, self.model = train_and_score(
self.network, dataset)
self.fit = 1 / self.loss if self.loss != 0 else sys.maxint
else:
print(
' Using Network (%2d/%2d) %s - accuracy: %.2f%%, - loss: %.4f'
% (count, total, self.print(), self.accuracy * 100, self.loss))
def train(self, x_train, y_train, x_test, y_test):
self.accuracy = train_and_score(self.network, x_train, y_train, x_test,
y_test)
def train(self, nb_classes, x, y):
if self.accuracy == 0.:
self.accuracy = train_and_score(self.network, nb_classes, x, y)
def train(self, dataset):
if self.accuracy == 0.:
self.accuracy = train_and_score(self.network, dataset)
def train(self, X, Y, generation):
self.accuracy = train_and_score(self.network, X, Y)
def train(self, data):
"""Train the network and record the loss."""
if self.loss == 10.:
self.loss = train_and_score(self.network, data)
def train(self, path, output):
self.loss = train_and_score(self.network, path, output)
name = path[-11:-3]
self.name = name
