def return_function(self, name):
dataset = dir_path.split('/')[1]
self.label = name + '-OCR-noisy'
if recordsExists(self.file, dataset, self.label, self.severity, self.noise_type):
return
status, score = getattr(self, 'if_' + name)()
compute_time = time() - self.start
bag = (self.file, dataset, self.label, self.severity, self.noise_type, status, score, compute_time, self.actual_str, self.detected_str)
writeToDB(bag)
x_test_max.shape[0], MAX_TESTING)
print("Number of Training Samples: X={}, Y={}".format(
x_train.shape[0], y_train.shape[0]))
print("Number of Validation Samples: X={}, Y={}".format(
x_val.shape[0], y_val.shape[0]))
print("Number of Test Samples: X={}, Y={}".format(x_test.shape[0],
y_test.shape[0]))
testing_optimizers = ['RAdam', 'Adam', 'NAdam', 'SGD']
testing_activations = ['tanh', 'relu', 'selu']
for optimizer in testing_optimizers:
for activation in testing_activations:
if recordsExists((architecture, label, optimizer, activation)):
print(
"Continuing for Optimizer = {} and Activation = {}".format(
optimizer, activation))
continue
loss, accuracy, mae, mse = {}, {}, {}, {}
train_loss_dict, train_accuracy_dict, train_mae_dict, train_mse_dict = {}, {}, {}, {}
if optimizer == 'RAdam':
score_returned, history = diff_optimizer(RAdam(lr=0.05),
activation,
label='RAdam')
else:
score_returned, history = diff_optimizer(optimizer, activation)