Revisiting Reweighted Wake-Sleep for Models with Stochastic Control Flow
This addresses a problem for researchers and practitioners in machine learning working with generative models that have stochastic control flow, offering a competitive alternative to existing methods, though it is incremental as it revisits an older algorithm.
The paper tackled the challenge of amortized gradient-based learning for stochastic control-flow models (SCFMs), which involve discrete random variables and branching paths, by revisiting the reweighted wake-sleep (RWS) algorithm. The result showed that RWS outperforms current state-of-the-art methods in learning SCFMs and learns better models and inference networks with increasing numbers of particles.
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 naive estimators. Here, we revisit the reweighted wake-sleep (RWS) (Bornschein and Bengio, 2015) 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.