Affiliation:
1. Department of Computer Science University of Toronto Toronto, Canada M5S lA7
Abstract
Most numerical techniques that have been proposed for the solution of constant-coefficient linear prob- Zems involve -either directly or indirectly-approxi mating a matrix exponential. For large systems these techniques can be very expensive to implement on digital computers. In this paper we describe how standard stiff ordinary differential equation (ODE) methods can be modified to take advantage of linear ity and thereby efficiently solve large linear prob lems. Methods based on backward differentiation formulas or second derivative formulas are particular ly suitable for this purpose. The implementation of these methods is discussed and numerical results are presented.
Subject
Computer Graphics and Computer-Aided Design,Modelling and Simulation,Software
Cited by
9 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献