Computing Reviews

Using AD to solve BVPs in MATLAB
Shampine L., Ketzscher R., Forth S. ACM Transactions on Mathematical Software31(1):79-94,2005.Type:Article
Date Reviewed: 04/20/05

A method for the numerical solution of two-point boundary value problems for ordinary differential equations, using MATLAB, is discussed in this paper. A classical algorithm for this problem has been proposed by Kierzenka and Shampine [1]. This method, known as bvp4c, works under very general assumptions. It is based on collocation techniques, and, for its implementation, one needs to compute many derivatives of the given functions. The original algorithm calculated these derivatives by a purely numerical method (finite differences). It turns out, however, that bvp4c can be substantially improved with respect to robustness and speed if these numerical derivatives are replaced by analytical ones. Thus, this paper discusses how such analytical derivatives can be provided for use with bvp4c in an efficient way. The method of choice is automatic differentiation. The resulting algorithm is described in detail. Illustrative examples are given, and the effect of vectorization is explained.


1)

Kierzenka, J.; Shampine, L. A BVP solver based on residual control and the MATLAB PSE. ACM Transact. Math. Software 27, (2001), 299–316.

Reviewer:  Kai Diethelm Review #: CR131164 (0510-1153)

Reproduction in whole or in part without permission is prohibited.   Copyright 2024 ComputingReviews.com™
Terms of Use
| Privacy Policy