Author
Le, T
Kosiorek, A
Siddharth, N
Teh, Y
Wood, F
Journal title
35th Conference on Uncertainty in Artificial Intelligence, UAI 2019
Last updated
2022-03-08T04:40:29.957+00:00
Abstract
© 2019 Association For Uncertainty in Artificial Intelligence (AUAI). All rights reserved. Stochastic control-flow models (SCFMs) are a class of generative models that involve branching on choices from discrete random variables. Amortized gradient-based learning of SCFMs is challenging as most approaches targeting discrete variables rely on their continuous relaxations—which can be intractable in SCFMs, as branching on relaxations requires evaluating all (exponentially many) branching paths. Tractable alternatives mainly combine REINFORCE with complex control-variate schemes to improve the variance of naïve estimators. Here, we revisit the reweighted wake-sleep (RWS) [5] algorithm, and through extensive evaluations, show that it outperforms current state-of-the-art methods in learning SCFMs. Further, in contrast to the importance weighted autoencoder, we observe that RWS learns better models and inference networks with increasing numbers of particles. Our results suggest that RWS is a competitive, often preferable, alternative for learning SCFMs.
Symplectic ID
895438
Publication type
Conference Paper
Publication date
1 January 2019
Please contact us with feedback and comments about this page. Created on 12 Aug 2018 - 17:30.