{"id":"arxiv/2608.29014","url":"https://arxiv.org/abs/2608.29014","tags":["math.NA"],"score":0,"title":"Gappy probabilistic manifold decomposition for nonlinear field reconstruction","domain":"math.NA","source":"arxiv","authors":["Qihan Feng","Jiaming Guo","Jiarun Meng","Dunhui Xiao"],"reviews":0,"subject":"math.NA","summary":"This paper proposes gappy probabilistic manifold decomposition (Gappy PMD), a nonlinear method for reconstructing high-dimensional fields from extremely sparse measurements. Gappy PMD reconstructs the field on the nonlinear manifold learned by probabilistic manifold decomposition (PMD). We further propose a differentiable point selection method for reduced-order model (ROM)-based field reconstruction (DPS). Using differentiable meshless interpolation within the ROM-based reconstruction framework, DPS makes the full-field reconstruction error differentiable with respect to the sampling…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-08-29","reactions":{}}