Open Peer Review. Open Publishing. Open Access. Open Discussion. Open Directory. Open Recommendations. Open API. Open Source.
Rethinking Style and Content Disentanglement in Variational Autoencoders
Rui Shu, Shengjia Zhao, Mykel J. Kochenderfer
Feb 12, 2018 (modified: Feb 12, 2018)ICLR 2018 Workshop Submissionreaders: everyone
Abstract:A common test for whether a generative model learns disentangled representations is its ability to learn style and content as independent factors of variation on digit datasets. To achieve such disentanglement with variational autoencoders, the label information is often provided in either a fully-supervised or semi-supervised fashion. We show, however, that the variational objective is insufficient in explaining the observed style and content disentanglement. Furthermore, we present an empirical framework to systematically evaluate the disentanglement behavior of our models. We show that the encoder and decoder independently favor disentangled representations and that this tendency depends on the implicit regularization by stochastic gradient descent.
TL;DR:Understanding deep representation learning requires rethinking disentanglement.
Keywords:disentangled representation, variational autoencoders, deep representation prior
Enter your feedback below and we'll get back to you as soon as possible.