Steepest descent and conjugate gradient methods with variable preconditioning
arXiv:math/0605767 · doi:10.1137/060675290
Abstract
We analyze the conjugate gradient (CG) method with variable preconditioning for solving a linear system with a real symmetric positive definite (SPD) matrix of coefficients . We assume that the preconditioner is SPD on each step, and that the condition number of the preconditioned system matrix is bounded above by a constant independent of the step number. We show that the CG method with variable preconditioning under this assumption may not give improvement, compared to the steepest descent (SD) method. We describe the basic theory of CG methods with variable preconditioning with the emphasis on ``worst case' scenarios, and provide complete proofs of all facts not available in the literature. We give a new elegant geometric proof of the SD convergence rate bound. Our numerical experiments, comparing the preconditioned SD and CG methods, not only support and illustrate our theoretical findings, but also reveal two surprising and potentially practically important effects. First, we analyze variable preconditioning in the form of inner-outer iterations. In previous such tests, the unpreconditioned CG inner iterations are applied to an artificial system with some fixed preconditioner as a matrix of coefficients. We test a different scenario, where the unpreconditioned CG inner iterations solve linear systems with the original system matrix . We demonstrate that the CG-SD inner-outer iterations perform as well as the CG-CG inner-outer iterations in these tests. Second, we show that variable preconditioning may surprisingly accelerate the SD and thus the CG convergence. Specifically, we compare the CG methods using a two-grid preconditioning with fixed and randomly chosen coarse grids, and observe that the fixed preconditioner method is twice as slow as the method with random preconditioning.
arXiv comment14 pages, 3 figures. Submitted to SIMAX
Cited by in corpus (9)
- Preconditioned Steepest Descent Methods for some Nonlinear Elliptic Equations Involving p-Laplacian Terms
- Nonsymmetric multigrid preconditioning for conjugate gradient methods
- Coarse-to-Fine Annotation Enrichment for Semantic Segmentation Learning
- Faster Least Squares Optimization
- Superfast CUR Matrix Algorithms, Their Pre-Processing and Extensions
- A unified convergence bound for conjugate gradient and accelerated gradient
- Preconditioned steepest descent-like methods for symmetric indefinite systems
- Accelerated graph-based spectral polynomial filters
- Adaptive Iterative Hessian Sketch via A-Optimal Subsampling