Fine-tuning a neural preconditioner with a differentiable FGMRES loss on principal angles cuts average FGMRES iterations roughly tenfold on mixed-dimensional 3D-1D problems.
Elsevier, Amsterdam, The Netherlands (2000)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
math.NA 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Neural Preconditioning via Krylov Subspace Geometry
Fine-tuning a neural preconditioner with a differentiable FGMRES loss on principal angles cuts average FGMRES iterations roughly tenfold on mixed-dimensional 3D-1D problems.