<ul><li>Large linear systems in computational science often use iterative solvers with preconditioners.</li><li>A novel approach using graph neural networks (GNNs) for preconditioner construction outperforms classical methods and neural network-based preconditioning.</li><li>The approach involves learning correction for well-established preconditioners from linear algebra with GNNs.</li><li>Extensive experiments demonstrate the superiority of the proposed approach and loss function over classical preconditioners.</li></ul>

Teaching Old Preconditioners New Tricks: How GNNs Supercharge Linear Solvers

Discover more