def train_encoder_classifier(self, alpha=0):
encoder_classifier, history = self.fit_encoder_classifier(alpha=alpha)
plots.loss(history,
self.image_directory,
name="losses_encoder_classifier")
self.save_model_weights(encoder_classifier, 'encoder_classifier')
return encoder_classifier
def train(self):
"""
Begin logging, train the autoencoder, use the autoencoder's history to plot loss curves, and save the parameters of the autoencoder, encoder, and decoder (respectively) to .h5 files.
:return: None
"""
if self.enable_logging:
logs.begin_logging(self.experiment_directory)
auto_encoder, encoder, decoder, history = self.fit_autoencoder()
plots.loss(history, self.image_directory)
self.save_model_weights(auto_encoder, encoder, decoder)
self.plot_results((encoder, decoder))
def train_contrastive_mlp(self):
"""
Begin logging, train the contrastive MLP, use the it's history to plot loss curves, and save the parameters
to .h5 files.
:return: None
"""
if self.enable_logging:
logs.begin_logging(self.directory)
contrastive_mlp, history = self.fit_contrastive_mlp()
plots.loss(history, self.image_directory)
self.save_model_weights(contrastive_mlp, "contrastive_mlp")
self.plot_logits(contrastive_mlp)
def train_classifier(self, alpha=0, evaluate=True, report=True):
if self.enable_logging:
logs.begin_logging(self.experiment_directory)
classifier, history = self.fit_classifier(alpha=alpha)
self.print_settings()
plots.loss(history, self.image_directory)
plots.accuracy(history, self.image_directory)
self.save_model_weights(classifier, 'classifier')
if evaluate:
print('Evaluation')
result = classifier.evaluate(self.x_test, self.y_test_binary,
batch_size=self.batch_size, )
print(result)
if evaluate and report:
filepath = os.path.abspath(os.path.join(self.experiment_directory, 'report.txt'))
report = open(filepath, 'w+')
report.write(f'Loss: {result[0]}\nAccuracy: {result[1]}')
report.close()
return classifier
def train(self):
"""
Begin logging, train the autoencoder, use the autoencoder's history to plot loss curves, and save the parameters
of the autoencoder, encoder, and decoder (respectively) to .h5 files.
:return: None
"""
t0 = time()
if self.enable_logging:
logs.begin_logging(self.experiment_directory)
auto_encoder, encoder, decoder, history = self.fit_autoencoder()
t1 = time()
t = t1 - t0
print(f"Variational autoencoder trained in {t} seconds.\n")
self.print_settings()
plots.loss(history, self.image_directory)
self.save_model_weights(auto_encoder, 'auto_encoder')
self.save_model_weights(encoder, 'encoder')
self.save_model_weights(decoder, 'decoder')
self.plot_results((encoder, decoder))
self.report_latent_space_classifiers(encoder)
return auto_encoder, encoder, decoder
def train(self):
"""
Begin logging, train the autoencoder, use the autoencoder's history to plot loss curves, and save_all the parameters
of the autoencoder, encoder, and decoder (respectively) to .h5 files.
:return: None
"""
t0 = time()
if self.enable_logging:
logs.begin_logging(self.experiment_directory)
auto_encoder, encoder, decoder, history = self.fit_autoencoder()
t1 = time()
t = t1 - t0
print(f"Variational autoencoder trained in {t} seconds.\n")
plots.loss(history, self.image_directory)
self.save_model_weights(auto_encoder, 'auto_encoder')
self.save_model_weights(encoder, 'encoder')
self.save_model_weights(decoder, 'decoder')
#self.save_prediction(encoder, data_filename='x_train_latent.npy')
self.save_prediction(encoder,
data=[self.gaussian_test, self.x_test],
data_filename='x_test_latent.npy')
#self.save_prediction(auto_encoder, data_filename='x_train_predict.npy')
self.save_prediction(auto_encoder,
data=[self.gaussian_test, self.x_test],
data_filename='x_test_predict.npy')
self.save_experiment_settings()
x_test_pred_file = os.path.abspath(
os.path.join(self.experiment_directory, 'x_test_predict.npy'))
x_test_pred = np.load(x_test_pred_file)
self.save_input_output_comparison(self.x_test, x_test_pred)
self.report_latent_space_classifiers(encoder)
self.plot_results((encoder, decoder))
return auto_encoder, encoder, decoder
