Neural Bayesian Filtering

Christopher Solinas, Radovan Haluška, David Sychrovský, Finbarr Timbers, Nolan Bard, Michael Buro, Martin Schmid, Nathan R. Sturtevant, Michael Bowling

Keywords: Partially observable systems, belief state modeling, particle filtering, bayesian filtering, normalizing flows

TL;DR: Neural Bayesian Filtering embeds beliefs and tracks multimodal posteriors by filtering in the embedding space.

Abstract:

We present Neural Bayesian Filtering (NBF), an algorithm for maintaining posteriors, called beliefs, over hidden states in partially observable systems.
NBF is trained to find a good latent representation of the beliefs induced by a task.
It maps beliefs to fixed-length embedding vectors, which can condition generative models for sampling.
During filtering, particle-style updates compute posteriors in this embedding space using incoming observations and environment dynamics.
NBF combines the computational efficiency of classical filters with the expressiveness of deep generative models - tracking rapidly shifting, multimodal beliefs while mitigating the risk of *particle impoverishment*.
We validate NBF in state estimation tasks in partially observable variants of Gridworld and the card game Goofspiel.

Primary Area: Probabilistic methods (e.g., variational inference, causal inference, Gaussian processes)

Submission Number: 9808