Salvato in:
Dettagli Bibliografici
Autori principali: Zeng, Zhijun, Gan, Weiye, Chen, Junqing, Shi, Zuoqiang
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2509.19816
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866908556585861120
author Zeng, Zhijun
Gan, Weiye
Chen, Junqing
Shi, Zuoqiang
author_facet Zeng, Zhijun
Gan, Weiye
Chen, Junqing
Shi, Zuoqiang
contents In many engineering and applied science domains, high-dimensional nonlinear filtering is still a challenging problem. Recent advances in score-based diffusion models offer a promising alternative for posterior sampling but require repeated retraining to track evolving priors, which is impractical in high dimensions. In this work, we propose the Conditional Score-based Filter (CSF), a novel algorithm that leverages a set-transformer encoder and a conditional diffusion model to achieve efficient and accurate posterior sampling without retraining. By decoupling prior modeling and posterior sampling into offline and online stages, CSF enables scalable score-based filtering across diverse nonlinear systems. Extensive experiments on benchmark problems show that CSF achieves superior accuracy, robustness, and efficiency across diverse nonlinear filtering scenarios.
format Preprint
id arxiv_https___arxiv_org_abs_2509_19816
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle An Efficient Conditional Score-based Filter for High Dimensional Nonlinear Filtering Problems
Zeng, Zhijun
Gan, Weiye
Chen, Junqing
Shi, Zuoqiang
Machine Learning
93E11, 62F15, 68T07
In many engineering and applied science domains, high-dimensional nonlinear filtering is still a challenging problem. Recent advances in score-based diffusion models offer a promising alternative for posterior sampling but require repeated retraining to track evolving priors, which is impractical in high dimensions. In this work, we propose the Conditional Score-based Filter (CSF), a novel algorithm that leverages a set-transformer encoder and a conditional diffusion model to achieve efficient and accurate posterior sampling without retraining. By decoupling prior modeling and posterior sampling into offline and online stages, CSF enables scalable score-based filtering across diverse nonlinear systems. Extensive experiments on benchmark problems show that CSF achieves superior accuracy, robustness, and efficiency across diverse nonlinear filtering scenarios.
title An Efficient Conditional Score-based Filter for High Dimensional Nonlinear Filtering Problems
topic Machine Learning
93E11, 62F15, 68T07
url https://arxiv.org/abs/2509.19816