Preventing posterior collapse in variational autoencoders for text generation via decoder regularization

10/28/2021

∙

by   Alban Petit, et al.

∙

0

∙

share

Variational autoencoders trained to minimize the reconstruction error are sensitive to the posterior collapse problem, that is the proposal posterior distribution is always equal to the prior. We propose a novel regularization method based on fraternal dropout to prevent posterior collapse. We evaluate our approach using several metrics and observe improvements in all the tested configurations.