Variational loss

Specifies an evaluation function adapted to the variational autoencoder. It combines a base reconstruction error and the Kullback-Leibler divergence between the learned distribution and the true latent posterior.

loss_variational(reconstruction_loss)

Arguments

reconstruction_loss

Another loss to be used as reconstruction error (e.g. "binary_crossentropy")

Value

A "ruta_loss" object

References

• Auto-Encoding Variational Bayes

• Under the Hood of the Variational Autoencoder (in Prose and Code)

• Keras example: Variational autoencoder

See also

\link{autoencoder_variational}

Other loss functions: contraction(), correntropy()