def fit(self, x_train, y_train, x_val, y_val):
if self.verbose > 0:
print("[{}] Fitting the regressor with data of {}".format(
self.name, x_train.shape))
start_time = time.time()
self.model = self.model.fit(x_train, y_train)
y_pred, _ = self.predict(x_train)
df_metrics = calculate_regression_metrics(y_train, y_pred)
self.train_duration = time.time() - start_time
df_metrics["duration"] = self.train_duration
save_train_duration(self.output_directory + 'train_duration.csv',
self.train_duration)
self.train_metrics = df_metrics
if self.verbose > 0:
print("[{}] Fitting completed, took {}s".format(
self.name, self.train_duration))
print("[{}] Fitting completed, best RMSE={}, MAE={}".format(
self.name, df_metrics["rmse"][0], df_metrics["mae"][0]))
return df_metrics
def fit(self, x_train, y_train, x_val, y_val):
if self.verbose > 0:
print("[{}] Fitting the regressor with data of {}".format(self.name, x_train.shape))
batch_size = 16
nb_epochs = 2000
mini_batch_size = int(min(x_train.shape[0] / 10, batch_size))
start_time = time.time()
hist = self.model.fit(x_train, y_train, batch_size=mini_batch_size, epochs=nb_epochs,
verbose=self.verbose, validation_data=(x_val, y_val), callbacks=self.callbacks)
self.train_duration = time.time() - start_time
y_pred, _ = self.predict(x_train)
df_metrics = calculate_regression_metrics(y_train, y_pred)
df_metrics["duration"] = self.train_duration
save_train_duration(self.output_directory + 'train_duration.csv', self.train_duration)
self.train_metrics = df_metrics
self.model.save(self.output_directory + 'last_model.hdf5')
save_logs_for_regression_deep_learning(self.output_directory, hist)
keras.backend.clear_session()
if self.verbose > 0:
print("[{}] Fitting completed, took {}s".format(self.name, self.train_duration))
print("[{}] Fitting completed, RMSE={}, MAE={}".format(self.name,
df_metrics["rmse"][0],
df_metrics["mae"][0]))
return df_metrics
def fit(self, x_train: np.array, y_train: np.array):
"""
Fit Linear model
Inputs:
x_train: training data (num_examples, num_timestep, num_channels) or (num_examples, num_features)
y_train: training target
"""
print("[{}] Fitting the regressor with data of {}".format(
self.name, x_train.shape))
start_time = time.perf_counter()
if len(x_train.shape) == 3:
x_train = x_train.reshape(x_train.shape[0],
x_train.shape[1] * x_train.shape[2])
self.cv(x_train, y_train)
self.model.fit(x_train, y_train)
self.train_duration = time.perf_counter() - start_time
save_train_duration(self.output_directory + 'train_duration.csv',
self.train_duration)
print("[{}] Fitting completed, took {}s".format(
self.name, self.train_duration))
def fit(self, x_train, y_train, x_val=None, y_val=None, monitor_val=False):
print('[{}] Training'.format(self.name))
start_time = time.perf_counter()
self.X_train = x_train
self.y_train = y_train
self.X_val = x_val
self.y_val = y_val
epochs = self.epochs
batch_size = self.batch_size
mini_batch_size = int(min(x_train.shape[0] / 10, batch_size))
file_path = self.output_directory + self.best_model_file
if (x_val is not None) and monitor_val:
reduce_lr = tf.keras.callbacks.ReduceLROnPlateau(
monitor='val_loss', factor=0.5, patience=50, min_lr=0.0001)
model_checkpoint = tf.keras.callbacks.ModelCheckpoint(
filepath=file_path, monitor='val_loss', save_best_only=True)
else:
reduce_lr = tf.keras.callbacks.ReduceLROnPlateau(monitor='loss',
factor=0.5,
patience=50,
min_lr=0.0001)
model_checkpoint = tf.keras.callbacks.ModelCheckpoint(
filepath=file_path, monitor='loss', save_best_only=True)
self.callbacks = [reduce_lr, model_checkpoint]
# train the model
if x_val is not None:
self.hist = self.model.fit(x_train,
y_train,
validation_data=(x_val, y_val),
verbose=self.verbose,
epochs=epochs,
batch_size=mini_batch_size,
callbacks=self.callbacks)
else:
self.hist = self.model.fit(x_train,
y_train,
verbose=self.verbose,
epochs=epochs,
batch_size=mini_batch_size,
callbacks=self.callbacks)
self.train_duration = time.perf_counter() - start_time
save_train_duration(self.output_directory + 'train_duration.csv',
self.train_duration)
print('[{}] Training done!, took {}s'.format(self.name,
self.train_duration))
def fit(self, x_train, y_train):
print("[{}] Fitting the regressor with data of {}".format(self.name, x_train.shape))
start_time = time.perf_counter()
if len(x_train.shape) == 3:
x_train = x_train.reshape(x_train.shape[0], x_train.shape[1] * x_train.shape[2])
self.cv(x_train, y_train)
self.model.fit(x_train, y_train)
self.train_duration = time.perf_counter() - start_time
save_train_duration(self.output_directory + 'train_duration.csv', self.train_duration)
print("[{}] Fitting completed, took {}s".format(self.name, self.train_duration))
def fit(self, x_train, y_train, x_val=None, y_val=None):
start_time = time.perf_counter()
print('[{}] Generating kernels'.format(self.name))
self.kernels = generate_kernels(x_train.shape[1], self.n_kernels,
x_train.shape[2])
print('[{}] Applying kernels'.format(self.name))
x_training_transform = apply_kernels(x_train, self.kernels)
print('[{}] Training'.format(self.name))
self.regressor.fit(x_training_transform, y_train)
self.train_duration = time.perf_counter() - start_time
save_train_duration(self.output_directory + 'train_duration.csv',
self.train_duration)
print('[{}] Training done!, took {}s'.format(self.name,
self.train_duration))
def fit(self, x_train, y_train, x_val, y_val):
# if len(keras.backend.tensorflow_backend._get_available_gpus()) == 0:
# print('error no gpu')
# exit()
if self.verbose > 0:
print("[{}] Fitting the regressor with data of {}".format(
self.name, x_train.shape))
# x_val and y_val are only used to monitor the test loss and NOT for training
if self.batch_size is None:
mini_batch_size = int(min(x_train.shape[0] / 10, 16))
else:
mini_batch_size = self.batch_size
start_time = time.time()
hist = self.model.fit(x_train,
y_train,
batch_size=mini_batch_size,
epochs=self.nb_epochs,
verbose=self.verbose,
validation_data=(x_val, y_val),
callbacks=self.callbacks)
self.train_duration = time.time() - start_time
y_pred, _ = self.predict(x_train)
df_metrics = calculate_regression_metrics(y_train, y_pred)
df_metrics["duration"] = self.train_duration
save_train_duration(self.output_directory + 'train_duration.csv',
self.train_duration)
self.train_metrics = df_metrics
self.model.save(self.output_directory + 'last_model.hdf5')
save_logs_for_regression_deep_learning(self.output_directory, hist)
keras.backend.clear_session()
if self.verbose > 0:
print("[{}] Fitting completed, took {}s".format(
self.name, self.train_duration))
print("[{}] Fitting completed, RMSE={}, MAE={}".format(
self.name, df_metrics["rmse"][0], df_metrics["mae"][0]))
return df_metrics
def fit(self, x_train, y_train, x_val=None, y_val=None, monitor_val=False):
"""
Fit DL models
Inputs:
x_train: training data (num_examples, num_timestep, num_channels)
y_train: training target
x_val: validation data (num_examples, num_timestep, num_channels)
y_val: validation target
monitor_val: boolean indicating if model selection should be done on validation
"""
print('[{}] Training'.format(self.name))
start_time = time.perf_counter()
self.X_train = x_train
self.y_train = y_train
self.X_val = x_val
self.y_val = y_val
epochs = self.epochs
batch_size = self.batch_size
mini_batch_size = int(min(x_train.shape[0] / 10, batch_size))
file_path = self.output_directory + self.best_model_file
if (x_val is not None) and monitor_val:
reduce_lr = tf.keras.callbacks.ReduceLROnPlateau(monitor='val_loss',
factor=0.5, patience=50,
min_lr=0.0001)
model_checkpoint = tf.keras.callbacks.ModelCheckpoint(filepath=file_path,
monitor='val_loss',
save_best_only=True)
else:
reduce_lr = tf.keras.callbacks.ReduceLROnPlateau(monitor='loss',
factor=0.5, patience=50,
min_lr=0.0001)
model_checkpoint = tf.keras.callbacks.ModelCheckpoint(filepath=file_path,
monitor='loss',
save_best_only=True)
self.callbacks = [reduce_lr, model_checkpoint]
# train the model
if x_val is not None:
self.hist = self.model.fit(x_train, y_train,
validation_data=(x_val, y_val),
verbose=self.verbose,
epochs=epochs,
batch_size=mini_batch_size,
callbacks=self.callbacks)
else:
self.hist = self.model.fit(x_train, y_train,
verbose=self.verbose,
epochs=epochs,
batch_size=mini_batch_size,
callbacks=self.callbacks)
self.train_duration = time.perf_counter() - start_time
save_train_duration(self.output_directory + 'train_duration.csv', self.train_duration)
print('[{}] Training done!, took {}s'.format(self.name, self.train_duration))
plot_epochs_metric(self.hist,
self.output_directory + 'epochs_loss.png',
metric='loss',
model=self.name)
for m in self.metrics:
plot_epochs_metric(self.hist,
self.output_directory + 'epochs_{}.png'.format(m),
metric=m,
model=self.name)