{"ID":2881284,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2508.13313","arxiv_id":"2508.13313","title":"Flow Matching for Efficient and Scalable Data Assimilation","abstract":"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.","short_abstract":"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...","url_abs":"https://arxiv.org/abs/2508.13313","url_pdf":"https://arxiv.org/pdf/2508.13313v3","authors":"[\"Taos Transue\",\"Bohan Chen\",\"So Takao\",\"Bao Wang\"]","published":"2025-08-18T19:00:45Z","proceeding":"stat.ML","tasks":"[\"stat.ML\",\"cs.LG\",\"math.OC\"]","methods":"[]","has_code":false,"code_links":[{"ID":610797,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_id":2881284,"paper_url":"https://arxiv.org/abs/2508.13313","paper_title":"Flow Matching for Efficient and Scalable Data Assimilation","repo_url":"https://github.com/Utah-Math-Data-Science/Data-Assimilation-Flow-Matching","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}