Factored Sparse Approximate Inverse Preconditioning via Spectral Optimization

DOI10.48550/arXiv.2606.22742MaRDI QIDQ7269913arXivFDO

Authors Tianshi Xu, Vassilis Kalantzis, Rui Peng Liu, Yuanzhe Xi, Francesco Brarda, Rui Peng Li

Publication date 22 June 2026

Published in arXiv

Tldr This work optimizes spectral objectives for selecting admissible entries in factorized sparse approximate inverse preconditioners using congruent operator metrics rather than classical Frobenius residual criteria. Projected Krylov gradients via detached Rayleigh surrogates enable practical implementation showing improved convergence on symmetric indefinite systems and offering a graph neural network model for entry prediction.

Full work available at URL https://doi.org/10.48550/arXiv.2606.22742

Mathematics Subject Classification ID

Preconditioners for iterative methods (65F08) Iterative numerical methods for linear systems (65F10) Artificial neural networks and deep learning (68T07) Machine vision and scene understanding (68T45)

arXiv classification

math.NA

This page was built for publication: Factored Sparse Approximate Inverse Preconditioning via Spectral Optimization

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7269913)