def binary_crossentropy(y_true, y_pred): return K.mean(K.binary_crossentropy(y_true, y_pred), axis=-1)
def binary_crossentropy(y_true, y_pred): return K.mean(K.binary_crossentropy(y_true, y_pred), axis=-1)
parser.add_argument("-w", "--weights", help=help_)
help_ = "Use mse loss instead of binary cross entropy (default)"
parser.add_argument("-m", "--mse", help=help_, action='store_true')
args = parser.parse_args()
vae, encoder, decoder, inputs, z_mean, z_log_var, latent_inputs, outputs = get_model(
input_shape, intermediate_dim, latent_dim)
models = (encoder, decoder)
data = (x_test, y_test)
# VAE loss = mse_loss or xent_loss + kl_loss
# inputs = K.flatten(inputs)
# outputs = K.flatten(outputs)
if args.mse:
reconstruction_loss = mse(inputs, outputs)
else:
reconstruction_loss = binary_crossentropy(inputs, outputs)
reconstruction_loss *= image_size * image_size
kl_loss = 1 + z_log_var - K.square(z_mean) - K.exp(z_log_var)
kl_loss = K.sum(kl_loss, axis=-1)
kl_loss *= -0.5
vae_loss = K.mean(reconstruction_loss + kl_loss)
vae.add_loss(vae_loss)
vae.compile(optimizer='adam')
vae.summary()
#plot_model(vae,
# to_file='vae_mlp.png',
# show_shapes=True)
train_data_gen = data_gen()
test_data_gen = data_gen('test')