adam = Adam(lr=lr, beta_1=0.9, beta_2=0.999, decay=0.0)
model.compile(loss=args.loss, optimizer=adam, metrics=['accuracy'])
output = model(a)
print("Featurizer: %s" % args.featurizer)
print("\n=================================")
print("Simple Neural Network Model")
print("=================================\n")
model.summary()
print("Kernel regularizer: %s %s" % (args.kernel_regularizer, args.regularizer_param))
print("Dropout rate: %s" % args.dropout_rate)
print("Loss function: %s" % args.loss)
print("Optimizer functions: %s" % args.optimizer)
print("Number of epochs: %s" % args.epochs)
print("Learning rate: %s" % lr)
print ()
model = model_func.nn_training_history(model, X, Y, args.epochs, args.batch_size)
general.check_path_exists(save_model_path)
model_func.save_model(model, save_model_path, neural_network=1)
print("\nExiting program.")
sys.stdout.close()
if method in ['simplenn']:
neural_network = True
if args.layer_dim == 3:
layers_dim = [2048, 128, 8, 4]
activation = ['relu', 'softmax', 'sigmoid']
elif args.layer_dim == 4:
layers_dim = [2048, 512, 128, 8, 4]
activation = ['relu', 'tanh', 'softmax', 'sigmoid']
elif args.layer_dim == 5:
layers_dim = [2048, 512, 128, 16, 8, 4]
activation = ['relu', 'tanh', 'softmax', 'tanh', 'sigmoid']
model = model_func.define_model(method, layers_dim, activation)
elif method in ['convnn']:
neural_network = method
feature_length = X.shape[1]
model = model_func.define_model(method, feature_length)
else:
model = model_func.define_model(method, model_param)
model = model_func.train_model(model,
num_split,
seed,
X,
Y,
neural_network=neural_network)
if save:
save_filepath = './saved_model/14_ecfp/' + "%s_%s_%s.h" % (
method, args.dataset[:args.dataset.rfind('.')], args.filename_append)
model_func.save_model(model, save_filepath, neural_network)
print("\nExiting program.")
layers_dim = [X.shape[1], 512, 128, 32, 8, Y.shape[1]]
activation = ['relu', 'relu', 'relu', 'relu', 'sigmoid']
epochs = args.epochs
print("Number of epochs:", epochs)
model_func.define_model(method)
model = model_func.build_simplenn_model(layers_dim=layers_dim,
activation=activation,
loss=args.loss,
optimizer=args.optimizer)
else:
epochs = 0
model = model_func.define_model(method, model_param)
model = model_func.train_model(model,
num_split,
seed,
X,
Y,
neural_network_epochs=epochs)
if args.save_model:
save_filepath = './saved_model/' + filename + '.h'
general.check_path_exists(save_filepath)
model_func.save_model(model, save_filepath, epochs)
print("\nExiting program.")
sys.stdout.close()