NPO, a transformer-based neural algebraic multigrid operator, learns to approximate the inverse of discretized PDE matrices and accelerates GMRES, though the evidence omits the AMG baseline and contains data inconsistencies.
Learn- ing neural PDE solvers with convergence guarantees
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
cs.CE 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
Neural Preconditioning Operator for Efficient PDE Solves
NPO, a transformer-based neural algebraic multigrid operator, learns to approximate the inverse of discretized PDE matrices and accelerates GMRES, though the evidence omits the AMG baseline and contains data inconsistencies.