Training-Free Bayesian Filtering with Generative Emulators

TL;DR

This paper introduces a training-free Bayesian filtering method using diffusion-based emulators, enabling scalable particle filtering in high-dimensional nonlinear dynamical systems without additional training.

Contribution

The authors propose a novel, training-free particle filtering approach leveraging diffusion-based emulators to improve scalability in high-dimensional systems.

Findings

Successfully applied to nonlinear chaotic systems including atmospheric dynamics.

Demonstrates scalability of particle filtering to high-dimensional settings.

Eliminates the need for additional training in Bayesian filtering.

Abstract

Bayesian filtering is a well-known problem that aims to estimate plausible states of a dynamical system from observations. Among existing approaches to solve this problem, particle filters are theoretically exact for non-linear dynamics and observations, but suffer from poor scalability in high dimensions. In this work, we show that diffusion-based emulators of dynamical systems can be used to implement, without additional training, an optimal variant of particle filters that has remained largely unexplored due to implementation challenges with classical numerical solvers. Experiments on nonlinear chaotic systems, including atmospheric dynamics, demonstrate that the proposed approach successfully scales particle filtering to high-dimensional settings.