{"id":"arxiv/2607.12613","url":"https://arxiv.org/abs/2607.12613","tags":["math.NA","math.OC"],"score":0,"title":"Tensor-Based Reduced-Order Modeling for Optimization-Based Inverse Problems","domain":"math.NA","source":"arxiv","authors":["Sahidul Islam","Andreas Mang","Maxim Olshanskii"],"reviews":0,"subject":"math.NA","summary":"We develop a tensor reduced-order modeling (TROM) framework for optimization-based inverse problems governed by parameter-dependent dynamical systems. The approach approximates the parameter-to-observation map directly in tensor-train (TT) format using TT-SVD or TT-Cross and integrates it into a regularized nonlinear least-squares formulation. Beyond accelerating forward evaluations, the low-rank tensor structure reformulates the inverse problem in reduced coordinates, assembles Gauss--Newton quantities without forming the full observation-space Jacobian, and minimizes the TROM objective…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-07-14","reactions":{}}