Hybrid Memoised Wake-Sleep: Approximate Inference at the Discrete-Continuous Interface

TL;DR

The paper introduces Hybrid Memoised Wake-Sleep (HMWS), a novel inference algorithm for hybrid discrete-continuous models that improves efficiency and accuracy by combining memoisation with learned recognition models, outperforming existing methods.

Contribution

It extends Memoised Wake-Sleep to effectively handle continuous variables through learned recognition models, enabling better inference in hybrid models.

Findings

HMWS outperforms state-of-the-art inference methods in experiments.

Effective handling of both discrete and continuous variables improves model performance.

Demonstrated success in GP-kernel learning and 3D scene understanding domains.

Abstract

Modeling complex phenomena typically involves the use of both discrete and continuous variables. Such a setting applies across a wide range of problems, from identifying trends in time-series data to performing effective compositional scene understanding in images. Here, we propose Hybrid Memoised Wake-Sleep (HMWS), an algorithm for effective inference in such hybrid discrete-continuous models. Prior approaches to learning suffer as they need to perform repeated expensive inner-loop discrete inference. We build on a recent approach, Memoised Wake-Sleep (MWS), which alleviates part of the problem by memoising discrete variables, and extend it to allow for a principled and effective way to handle continuous variables by learning a separate recognition model used for importance-sampling based approximate inference and marginalization. We evaluate HMWS in the GP-kernel learning and 3D scene understanding domains, and show that it outperforms current state-of-the-art inference methods.