{"id":"arxiv/2606.22742","url":"https://arxiv.org/abs/2606.22742","tags":["math.NA"],"score":0,"title":"Factored Sparse Approximate Inverse Preconditioning via Spectral Optimization","domain":"math.NA","source":"arxiv","authors":["Francesco Brarda","Tianshi Xu","Vassilis Kalantzis","Rui Peng Li","Yuanzhe Xi"],"reviews":0,"subject":"math.NA","summary":"In this paper, we study value selection for fixed-pattern factorized sparse approximate inverse preconditioners. Given a prescribed sparsity pattern for a factor $G,$ we choose its admissible entries by optimizing spectral objectives of the congruent preconditioned operator $P(G)=GAG^T.$ This differs from classical sparse approximate inverse and FSAI constructions, which choose entries through algebraic Frobenius-residual criteria. For symmetric positive definite systems, the spectral target is a cluster near $+1.$ For symmetric indefinite systems, where congruence preserves inertia, we…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-06-22","reactions":{}}