{"id":"arxiv/2610.13483","url":"https://arxiv.org/abs/2610.13483","tags":["math.NA"],"score":0,"title":"Regularisation and Parameter Choice Rules for Composite Linear Inverse Problems","domain":"math.NA","source":"arxiv","authors":["Christina Brandt","Patrick Horn"],"reviews":0,"subject":"math.NA","summary":"In this paper we propose a novel regularisation strategy for structured linear inverse problems of the form $Ax=F \\circ G x = z$. Instead of using the operator $A$ or a complex scheme involving both $F$ and $G$ simultaneously, combining already existing and simpler regularisations for $F$ and $G$ allows the use of established and optimised programs for high-speed computations. As classical parameter choice rules fail in the scenario when no source conditions are known because intermediate error bounds may be unavailable, we provide both a-priori and a-posteriori rules with a convergence…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-09","reactions":{}}