Pith. sign in

REVIEW

On Posterior Collapse and Encoder Feature Dispersion in Sequence VAEs

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 1911.03976 v2 pith:POQY2GKA submitted 2019-11-10 cs.LG cs.CLstat.ML

On Posterior Collapse and Encoder Feature Dispersion in Sequence VAEs

classification cs.LG cs.CLstat.ML
keywords collapseposteriorvaesmodelsequenceachievedispersionencoder
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Variational autoencoders (VAEs) hold great potential for modelling text, as they could in theory separate high-level semantic and syntactic properties from local regularities of natural language. Practically, however, VAEs with autoregressive decoders often suffer from posterior collapse, a phenomenon where the model learns to ignore the latent variables, causing the sequence VAE to degenerate into a language model. In this paper, we argue that posterior collapse is in part caused by the lack of dispersion in encoder features. We provide empirical evidence to verify this hypothesis, and propose a straightforward fix using pooling. This simple technique effectively prevents posterior collapse, allowing model to achieve significantly better data log-likelihood than standard sequence VAEs. Comparing to existing work, our proposed method is able to achieve comparable or superior performances while being more computationally efficient.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.