Gradient estimation for binary latent variables via gradient variance clipping Conference Paper
| Authors: | Kunes, R. Z.; Yin, M.; Land, M.; Haviv, D.; Peer, D.; Tavaré, S. |
| Title: | Gradient estimation for binary latent variables via gradient variance clipping |
| Conference Title: | 37th AAAI Conference on Artificial Intelligence (AAAI 2023) |
| Abstract: | Gradient estimation is often necessary for fitting generative models with discrete latent variables, in contexts such as reinforcement learning and variational autoencoder (VAE) training. The DisARM estimator achieves state of the art gradient variance for Bernoulli latent variable models in many contexts. However, DisARM and other estimators have potentially exploding variance near the boundary of the parameter space, where solutions tend to lie. To ameliorate this issue, we propose a new gradient estimator bitflip-1 that has lower variance at the boundaries of the parameter space. As bitflip-1 has complementary properties to existing estimators, we introduce an aggregated estimator, unbiased gradient variance clipping (UGC) that uses either a bitflip-1 or a DisARM gradient update for each coordinate. We theoretically prove that UGC has uniformly lower variance than DisARM. Empirically, we observe that UGC achieves the optimal value of the optimization objectives in toy experiments, discrete VAE training, and in a best subset selection problem. Copyright © 2023, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. |
| Keywords: | parameter estimation; learning systems; latent variable; state of the art; generative model; auto encoders; reinforcement learning; reinforcement learnings; bit-flips; gradient estimation; gradient variance; in contexts; parameter spaces; toys |
| Journal Title | Proceedings of the 37th AAAI Conference on Artificial Intelligence (AAAI 2023) |
| Volume: | 37 |
| Conference Dates: | 2023 Feb 7-14 |
| Conference Location: | Washington, D. C. |
| ISBN: | 978-1-57735-880-0 |
| Publisher: | Aaai Press |
| Date Published: | 2023-06-27 |
| Start Page: | 8405 |
| End Page: | 8412 |
| Language: | English |
| PROVIDER: | scopus |
| DOI: | 10.1609/aaai.v37i7.26013 |
| DOI/URL: | |
| Notes: | Conference paper -- Source: Scopus |
Altmetric
Citation Impact
BMJ Impact Analytics
Related MSK Work