What question did this study set out to answer?

To develop a novel preconditioner design using graph neural networks that improves the condition number in large linear systems.

Learning from Linear Algebra: A Graph Neural Network Approach to Preconditioner Design for Conjugate Gradient Solvers

View Full Paper

Ask AI

Bookmark

Authors

Demonstrates a new method for improving preconditioners in solving large linear systems, suggesting novel efficiency gains.

To develop a novel preconditioner design using graph neural networks that improves the condition number in large linear systems.
Utilized existing preconditioners from linear algebra as a starting point for training GNNs.
Conducted numerical experiments to compare the performance of GNN-based preconditioners against classical methods.
Employed a heuristic justification for the loss function used in training the GNN.
GNN-based preconditioners yielded a more substantial reduction in the condition number compared to classical preconditioners.
Numerical experiments showed improved efficiency for solving parametric PDEs with the new approach.
The loss function used contributed to more desirable reductions in condition numbers for the conjugate gradient method.

Trifonov et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark