Another Look at Neural Multigrid

DOI10.1142/S0129183197000187MaRDI QIDQ4488312zbMATH OpenOpenAlexFDO

Authors Martin Bäker

Publication date 5 July 2000

Published in International Journal of Modern Physics C (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/hep-lat/9610035

zbMATH Keywords

algorithmconjugate gradient methodgauge fieldsinterpolation operatorsneural multigrid methodneural netordered Laplace equation

Mathematics Subject Classification ID

Laplace operator, Helmholtz equation (reduced wave equation), Poisson equation (35J05) Iterative numerical methods for linear systems (65F10) Finite difference methods for boundary value problems involving PDEs (65N06) Multigrid methods; domain decomposition for boundary value problems involving PDEs (65N55) Quantum field theory on lattices (81T25) Neural nets applied to problems in time-dependent statistical mechanics (82C32)

Abstract: We present a new multigrid method called neural multigrid which is based on joining multigrid ideas with concepts from neural nets. The main idea is to use the Greenbaum criterion as a cost functional for the neural net. The algorithm is able to learn efficient interpolation operators in the case of the ordered Laplace equation with only a very small critical slowing down and with a surprisingly small amount of work comparable to that of a Conjugate Gradient solver. In the case of the two-dimensional Laplace equation with SU(2) gauge fields at beta=0 the learning exhibits critical slowing down with an exponent of about z = 0.4. The algorithm is able to find quite good interpolation operators in this case as well. Thereby it is proven that a practical true multigrid algorithm exists even for a gauge theory. An improved algorithm using dynamical blocks that will hopefully overcome the critical slowing down completely is sketched.

Recommendations

• Multilevel Artificial Neural Network Training for Spatially Correlated Learning
• Black-box learning of multigrid parameters
• Optimal multigrid algorithms for variable-coupling isotropic Gaussian models
• Accelerating algebraic multigrid methods via artificial neural networks
• Optimal multigrid algorithms for the massive Gaussian model and path integrals

Cites work

• Analysis of a Multigrid Method as an Iterative Technique for Solving Linear Systems

Cited in

(5)

• Black-box learning of multigrid parameters
• MgNet: a unified framework of multigrid and convolutional neural network
• Accelerating algebraic multigrid methods via artificial neural networks
• Non-Abelian projective multigrid for lattice gauge theory
• Connections between numerical algorithms for PDEs and neural networks

This page was built for publication: Another Look at Neural Multigrid

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4488312)