def train_optimal_action_given_history_obs(
model,
target_history,
target_stocks,
window_length,
weight_path='weights/imitation_3_stocks.h5'):
nb_classes = len(target_stocks) + 1
(X_train, y_train), (X_validation,
y_validation) = create_imitation_dataset(
target_history, window_length)
Y_train = np_utils.to_categorical(y_train, nb_classes)
Y_validation = np_utils.to_categorical(y_validation, nb_classes)
X_train = np.expand_dims(X_train, axis=-1)
X_validation = np.expand_dims(X_validation, axis=-1)
continue_train = True
while continue_train:
model.fit(X_train,
Y_train,
batch_size=128,
epochs=100,
validation_data=(X_validation, Y_validation),
shuffle=True)
save_weights = input('Type True to save weights\n')
if save_weights:
model.save(weight_path)
continue_train = input(
"True to continue train, otherwise stop training...\n")
required=True)
args = vars(parser.parse_args())
pprint.pprint(args)
assert args['predictor_type'] in ['CNN', 'LSTM']
window_length = int(args['window_length'])
model_path = 'weights/{}.h5'.format(
get_model_name(args['predictor_type'], window_length))
if args['predictor_type'] == 'CNN':
model = StockCNN(nb_classes, window_length, model_path)
else:
model = StockLSTM(nb_classes, window_length, model_path)
model.build_model(load_weights=False)
(X_train, y_train), (X_validation,
y_validation) = create_imitation_dataset(
target_history, window_length)
Y_train = np_utils.to_categorical(y_train, nb_classes)
Y_validation = np_utils.to_categorical(y_validation, nb_classes)
if args['predictor_type'] == 'CNN':
X_train = np.expand_dims(X_train, axis=-1)
X_validation = np.expand_dims(X_validation, axis=-1)
model.train(X_train, Y_train, X_validation, Y_validation)