Flow Matching for Efficient and Scalable Data Assimilation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Transue, Taos, Chen, Bohan, Takao, So, Wang, Bao
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912611114680320
author Transue, Taos
Chen, Bohan
Takao, So
Wang, Bao
author_facet Transue, Taos
Chen, Bohan
Takao, So
Wang, Bao
contents Data assimilation (DA) estimates a dynamical system's state from noisy observations. Recent generative models like the ensemble score filter (EnSF) improve DA in high-dimensional nonlinear settings but are computationally expensive. We introduce the ensemble flow filter (EnFF), a training-free, flow matching (FM)-based framework that accelerates sampling and offers flexibility in flow design. EnFF uses Monte Carlo estimators for the marginal flow field, localized guidance for observation assimilation, and utilizes a novel flow that exploits the Bayesian DA formulation. It generalizes classical filters such as the bootstrap particle filter and ensemble Kalman filter. Experiments on high-dimensional benchmarks demonstrate EnFF's improved cost-accuracy tradeoffs and scalability, highlighting FM's potential for efficient, scalable DA. Code is available at https://github.com/Utah-Math-Data-Science/Data-Assimilation-Flow-Matching.
format Preprint
id arxiv_https___arxiv_org_abs_2508_13313
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Flow Matching for Efficient and Scalable Data Assimilation
Transue, Taos
Chen, Bohan
Takao, So
Wang, Bao
Machine Learning
Optimization and Control
60G35 (Primary), 62M20 (Secondary), 93E11
Data assimilation (DA) estimates a dynamical system's state from noisy observations. Recent generative models like the ensemble score filter (EnSF) improve DA in high-dimensional nonlinear settings but are computationally expensive. We introduce the ensemble flow filter (EnFF), a training-free, flow matching (FM)-based framework that accelerates sampling and offers flexibility in flow design. EnFF uses Monte Carlo estimators for the marginal flow field, localized guidance for observation assimilation, and utilizes a novel flow that exploits the Bayesian DA formulation. It generalizes classical filters such as the bootstrap particle filter and ensemble Kalman filter. Experiments on high-dimensional benchmarks demonstrate EnFF's improved cost-accuracy tradeoffs and scalability, highlighting FM's potential for efficient, scalable DA. Code is available at https://github.com/Utah-Math-Data-Science/Data-Assimilation-Flow-Matching.
title Flow Matching for Efficient and Scalable Data Assimilation
topic Machine Learning
Optimization and Control
60G35 (Primary), 62M20 (Secondary), 93E11
url https://arxiv.org/abs/2508.13313