Learning Optimal Multigrid Smoothers via Neural Networks
From MaRDI portal
Norms of matrices, numerical range, applications of functional analysis to matrix theory (15A60) Preconditioners for iterative methods (65F08) Iterative numerical methods for linear systems (65F10) Multigrid methods; domain decomposition for initial value and initial-boundary value problems involving PDEs (65M55)
Abstract: Multigrid methods are one of the most efficient techniques for solving linear systems arising from Partial Differential Equations (PDEs) and graph Laplacians from machine learning applications. One of the key components of multigrid is smoothing, which aims at reducing high-frequency errors on each grid level. However, finding optimal smoothing algorithms is problem-dependent and can impose challenges for many problems. In this paper, we propose an efficient adaptive framework for learning optimized smoothers from operator stencils in the form of convolutional neural networks (CNNs). The CNNs are trained on small-scale problems from a given type of PDEs based on a supervised loss function derived from multigrid convergence theories, and can be applied to large-scale problems of the same class of PDEs. Numerical results on anisotropic rotated Laplacian problems demonstrate improved convergence rates and solution time compared with classical hand-crafted relaxation methods.
Recommendations
- Black-box learning of multigrid parameters
- Learning-based local weighted least squares for algebraic multigrid method
- Accelerating algebraic multigrid methods via artificial neural networks
- An enhanced V-cycle MgNet model for operator learning in numerical partial differential equations
- \(\mathrm{AutoAMG}(\theta)\): an auto-tuned AMG method based on deep learning for strong threshold
Cites work
- A comparative study of sparse approximate inverse preconditioners
- A machine learning framework for data driven acceleration of computations of differential equations
- A Multigrid Tutorial, Second Edition
- Aggregation-Based Algebraic Multigrid for Convection-Diffusion Equations
- Algebraic multigrid based on element interpolation (AMGe)
- Algebraic multigrid for Markov chains
- Algebraic multigrid methods
- Algebraic multigrid theory: The symmetric case
- Approximate inverse preconditionings for sparse linear systems
- DGM: a deep learning algorithm for solving partial differential equations
- General solutions for nonlinear differential equations: a rule-based self-learning approach using deep reinforcement learning
- scientific article; zbMATH DE number 3874483 (Why is no real title available?)
- scientific article; zbMATH DE number 1953444 (Why is no real title available?)
- MgNet: a unified framework of multigrid and convolutional neural network
- Multigrid smoothers for ultraparallel computing
- Multilevel aggregation methods for small-world graphs with application to random-walk ranking
- On Generalizing the Algebraic Multigrid Framework
- On the Robustness of ILU Smoothing
- Parallel Preconditioning with Sparse Approximate Inverses
- Preconditioning
- Sharp Estimates for Multigrid Rates of Convergence with General Smoothing and Acceleration
- Small-Sample Statistical Estimates for Matrix Norms
- SMASH: structured matrix approximation by separation and hierarchy.
- Solving high-dimensional partial differential equations using deep learning
- The explicit block relaxation method as a grid smoother in the multigrid v-cycle scheme
- Universality of deep convolutional neural networks
Cited in
(17)- Recurrent neural networks as optimal mesh refinement strategies
- Accelerating algebraic multigrid methods via artificial neural networks
- Learning-based local weighted least squares for algebraic multigrid method
- Machine learning optimization of compact finite volume methods on unstructured grids
- DNN-MG: a hybrid neural network/finite element method with applications to 3D simulations of the Navier-Stokes equations
- Automated linear solver selection for simulation of multiphysics processes in porous media
- Reducing operator complexity of Galerkin coarse-grid operators with machine learning
- An enhanced V-cycle MgNet model for operator learning in numerical partial differential equations
- Optimal polynomial smoothers for parallel AMG
- Momentum-accelerated Richardson(m) neural solvers and their multi-level extensions
- Optimized sparse matrix operations for reverse mode automatic differentiation
- A hybrid iterative neural solver based on spectral analysis for parametric PDEs
- A neural multigrid solver for Helmholtz equations with high wavenumber and heterogeneous media
- Mitigating distribution shift in machine learning-augmented hybrid simulation
- Optimizing coarse propagators in parareal algorithms
- FastAGMGar: an aggregation-based algebraic multigrid method
- DeepONet based preconditioning strategies for solving parametric linear systems of equations
This page was built for publication: Learning Optimal Multigrid Smoothers via Neural Networks
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6108149)
