Uncertainty-Calibrated Spatiotemporal Field Diffusion with Sparse Supervision
#spatiotemporal #diffusion models #uncertainty calibration #sparse supervision #forecasting
📌 Key Takeaways
- Researchers propose a new method for modeling spatiotemporal fields with sparse data.
- The approach uses diffusion models to generate accurate predictions with calibrated uncertainty.
- It addresses challenges in forecasting where data is limited or unevenly distributed.
- The method improves reliability in applications like weather prediction and environmental monitoring.
📖 Full Retelling
arXiv:2603.04431v1 Announce Type: cross
Abstract: Physical fields are typically observed only at sparse, time-varying sensor locations, making forecasting and reconstruction ill-posed and uncertainty-critical. We present SOLID, a mask-conditioned diffusion framework that learns spatiotemporal dynamics from sparse observations alone: training and evaluation use only observed target locations, requiring no dense fields and no pre-imputation. Unlike prior work that trains on dense reanalysis or si
🏷️ Themes
Machine Learning, Data Science
Entity Intersection Graph
No entity connections available yet for this article.
Original Source
--> Computer Science > Machine Learning arXiv:2603.04431 [Submitted on 17 Feb 2026] Title: Uncertainty-Calibrated Spatiotemporal Field Diffusion with Sparse Supervision Authors: Kevin Valencia , Xihaier Luo , Shinjae Yoo , David Keetae Park View a PDF of the paper titled Uncertainty-Calibrated Spatiotemporal Field Diffusion with Sparse Supervision, by Kevin Valencia and 3 other authors View PDF HTML Abstract: Physical fields are typically observed only at sparse, time-varying sensor locations, making forecasting and reconstruction ill-posed and uncertainty-critical. We present SOLID, a mask-conditioned diffusion framework that learns spatiotemporal dynamics from sparse observations alone: training and evaluation use only observed target locations, requiring no dense fields and no pre-imputation. Unlike prior work that trains on dense reanalysis or simulations and only tests under sparsity, SOLID is trained end-to-end with sparse supervision only. SOLID conditions each denoising step on the measured values and their locations, and introduces a dual-masking objective that emphasizes learning in unobserved void regions while upweights overlap pixels where inputs and targets provide the most reliable anchors. This strict sparse-conditioning pathway enables posterior sampling of full fields consistent with the measurements, achieving up to an order-of-magnitude improvement in probabilistic error and yielding calibrated uncertainty maps (\r > 0.7) under severe sparsity. Comments: 18 pages, 9 figures, 6 tables Subjects: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI) Cite as: arXiv:2603.04431 [cs.LG] (or arXiv:2603.04431v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2603.04431 Focus to learn more arXiv-issued DOI via DataCite Submission history From: David Keetae Park [ view email ] [v1] Tue, 17 Feb 2026 16:56:55 UTC (13,585 KB) Full-text links: Access Paper: View a PDF of the paper titled Uncertainty-Calibrated Spatiotemporal Field Diffusion with S...
Read full article at source