arrow
Volume 13, Issue 2
Adaptive Relaxation Strategy on Basic Iterative Methods for Solving Linear Systems with Single and Multiple Right-Hand Sides

Yuan Yuan, Shuli Sun, Pu Chen & Mingwu Yuan

Adv. Appl. Math. Mech., 13 (2021), pp. 378-403.

Published online: 2020-12

Export citation
  • Abstract

Two adaptive techniques for choosing relaxation factor, namely, Minimal Residual Relaxation (MRR) and Orthogonal Projection Relaxation (OPR), on basic iterative methods for solving linear systems are proposed. Unlike classic relaxation, in which the optimal relaxation factor is generally difficult to find, in these proposed techniques, non-stationary relaxation factor based on minimal residual or orthogonal projection method is calculated adaptively in each relaxation step with acceptable cost for Jacobi, Gauss-Seidel or symmetric Gauss-Seidel iterative methods. In order to avoid the "stagnation" of the successive locally optimal relaxations, a recipe of inserting several basic iterations between every two adjacent relaxations is suggested and the resulting MRR$(m)$/OPR$(m)$ strategy is more stable and efficient (here $m$ denotes the number of basic iterations inserted). To solve linear systems with multiple right-hand sides efficiently, block-form relaxation strategies are proposed based on the MRR$(m)$ and OPR$(m)$. Numerical experiments show that the presented MRR$(m)$/OPR$(m)$ algorithm is more robust and effective than classic relaxation methods. It is also showed that the proposed block relaxation strategies can efficiently accelerate the solution of systems with multiple right-hand sides in terms of total solution time as well as number of iterations.

  • AMS Subject Headings

65F10, 65B99

  • Copyright

COPYRIGHT: © Global Science Press

  • Email address
  • BibTex
  • RIS
  • TXT
@Article{AAMM-13-378, author = {Yuan , YuanSun , ShuliChen , Pu and Yuan , Mingwu}, title = {Adaptive Relaxation Strategy on Basic Iterative Methods for Solving Linear Systems with Single and Multiple Right-Hand Sides}, journal = {Advances in Applied Mathematics and Mechanics}, year = {2020}, volume = {13}, number = {2}, pages = {378--403}, abstract = {

Two adaptive techniques for choosing relaxation factor, namely, Minimal Residual Relaxation (MRR) and Orthogonal Projection Relaxation (OPR), on basic iterative methods for solving linear systems are proposed. Unlike classic relaxation, in which the optimal relaxation factor is generally difficult to find, in these proposed techniques, non-stationary relaxation factor based on minimal residual or orthogonal projection method is calculated adaptively in each relaxation step with acceptable cost for Jacobi, Gauss-Seidel or symmetric Gauss-Seidel iterative methods. In order to avoid the "stagnation" of the successive locally optimal relaxations, a recipe of inserting several basic iterations between every two adjacent relaxations is suggested and the resulting MRR$(m)$/OPR$(m)$ strategy is more stable and efficient (here $m$ denotes the number of basic iterations inserted). To solve linear systems with multiple right-hand sides efficiently, block-form relaxation strategies are proposed based on the MRR$(m)$ and OPR$(m)$. Numerical experiments show that the presented MRR$(m)$/OPR$(m)$ algorithm is more robust and effective than classic relaxation methods. It is also showed that the proposed block relaxation strategies can efficiently accelerate the solution of systems with multiple right-hand sides in terms of total solution time as well as number of iterations.

}, issn = {2075-1354}, doi = {https://doi.org/10.4208/aamm.OA-2019-0378}, url = {http://global-sci.org/intro/article_detail/aamm/18489.html} }
TY - JOUR T1 - Adaptive Relaxation Strategy on Basic Iterative Methods for Solving Linear Systems with Single and Multiple Right-Hand Sides AU - Yuan , Yuan AU - Sun , Shuli AU - Chen , Pu AU - Yuan , Mingwu JO - Advances in Applied Mathematics and Mechanics VL - 2 SP - 378 EP - 403 PY - 2020 DA - 2020/12 SN - 13 DO - http://doi.org/10.4208/aamm.OA-2019-0378 UR - https://global-sci.org/intro/article_detail/aamm/18489.html KW - Adaptive relaxation, basic iterative method, minimal residual, orthogonal projection, linear system with multiple right-hand sides. AB -

Two adaptive techniques for choosing relaxation factor, namely, Minimal Residual Relaxation (MRR) and Orthogonal Projection Relaxation (OPR), on basic iterative methods for solving linear systems are proposed. Unlike classic relaxation, in which the optimal relaxation factor is generally difficult to find, in these proposed techniques, non-stationary relaxation factor based on minimal residual or orthogonal projection method is calculated adaptively in each relaxation step with acceptable cost for Jacobi, Gauss-Seidel or symmetric Gauss-Seidel iterative methods. In order to avoid the "stagnation" of the successive locally optimal relaxations, a recipe of inserting several basic iterations between every two adjacent relaxations is suggested and the resulting MRR$(m)$/OPR$(m)$ strategy is more stable and efficient (here $m$ denotes the number of basic iterations inserted). To solve linear systems with multiple right-hand sides efficiently, block-form relaxation strategies are proposed based on the MRR$(m)$ and OPR$(m)$. Numerical experiments show that the presented MRR$(m)$/OPR$(m)$ algorithm is more robust and effective than classic relaxation methods. It is also showed that the proposed block relaxation strategies can efficiently accelerate the solution of systems with multiple right-hand sides in terms of total solution time as well as number of iterations.

Yuan Yuan, Shuli Sun, Pu Chen & Mingwu Yuan. (1970). Adaptive Relaxation Strategy on Basic Iterative Methods for Solving Linear Systems with Single and Multiple Right-Hand Sides. Advances in Applied Mathematics and Mechanics. 13 (2). 378-403. doi:10.4208/aamm.OA-2019-0378
Copy to clipboard
The citation has been copied to your clipboard