{"id":"arxiv/2609.37140","url":"https://arxiv.org/abs/2609.37140","tags":["math.NA","cs.LG"],"score":0,"title":"BandPC: Learning Residual-Band Preconditioner Combinations for Flexible Conjugate Gradient Solvers","domain":"math.NA","source":"arxiv","authors":["D. M. Li","Shang-Tian Yang","Xin Qiu"],"reviews":0,"subject":"math.NA","summary":"The conjugate gradient (CG) method is a classical iterative solver for sparse symmetric positive definite (SPD) linear systems, but its convergence strongly depends on the spectral properties of the system matrix. Preconditioning can improve these properties; however, designing preconditioners that generalize across diverse and irregular sparsity patterns remains challenging. We propose BandPC (Band Preconditioner Combinations, where \"Band\" denotes residual-norm bands), a data-driven framework that leverages graph neural networks (GNNs) to predict multi-stage preconditioning strategies for…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-09-29","reactions":{}}