Stick-Breaking Variational Autoencoders

Eric Nalisnick, Padhraic Smyth
5/20/2016
stat.ML

Official code: enalisnick/stick-breaking_dgms (95 stars, MIT, last commit 2016-05-23).

Abstract

We extend Stochastic Gradient Variational Bayes to perform posterior inference for the weights of Stick-Breaking processes. This development allows us to define a Stick-Breaking Variational Autoencoder (SB-VAE), a Bayesian nonparametric version of the variational autoencoder that has a latent representation with stochastic dimensionality. We experimentally demonstrate that the SB-VAE, and a semi-supervised variant, learn highly discriminative latent representations that often outperform the Gaussian VAE's.

View on arXivView PDF

Code Implementations(1)

enalisnick/stick-breaking_dgmsOfficial100%
MIT

Deep Generative Models with Stick-Breaking Priors

9519PythonMay 20, 201610 years agoMIT

Similar Papers

πVAE: a stochastic process prior for Bayesian deep learning with MCMC

85% match1 repo142020

Lagging Inference Networks and Posterior Collapse in Variational Autoencoders

83% match1 repo1852019

Entropic Auto-Encoding via Implicit Free-Energy Minimization

82% match2026

Posterior Collapse and Latent Variable Non-identifiability

82% match2023

Towards Model-Agnostic Posterior Approximation for Fast and Accurate Variational Autoencoders

81% match2024

Learning variational autoencoders via MCMC speed measures

81% match2023

Cite this paper

@article{nalisnick2016stickbreaking,
  title  = {Stick-Breaking Variational Autoencoders},
  author = {Eric Nalisnick and Padhraic Smyth},
  year   = {2016},
  eprint = {1605.06197},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1605.06197}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1605.06197.svg)](https://codeofpaper.com/paper/1605.06197)

Discussion