Computer Science Technical Reports
CS at VT

Globally Convergent Homotopy Methods: A Tutorial

Watson, Layne T. (1987) Globally Convergent Homotopy Methods: A Tutorial. Technical Report TR-87-13, Computer Science, Virginia Polytechnic Institute and State University.

Full text available as:
PDF - Requires Adobe Acrobat Reader or other PDF viewer.
TR-87-13.pdf (2097832)

Abstract

The basic theory for probability one globally convergent homotopy algorithms was developed in 1976, and since then the theory; algorithms, and applications have considerably expanded. These are algorithms for solving nonlinear systems of (algebraic) equations, which are convergent for almost all choices of starting point. Thus they are globally convergent with probability one. They are applicable to Brouwer fixed point problems, certain classes of @erofin mg problems unconstrained optimization, linearly constrained optimization, nonlinear complementarity, and the discretizations of nonlinear two-point boundary value problems based on shooting, finite differences, collocation, and finite elements. A mathematical software package, HOMPACK, exists that implements several different strategies and handles both dense and sparse problems. Homotopy algorithms are closely related to ODE algorithms, and make heavy use of ODE techniques. Homotopy algorithms for some classes of nonlinear systems, such as polynomial systems, exhibit a large amount of coarse grain parallelism. These and other topics are discussed in a tutorial fashion.

Item Type:Departmental Technical Report
Subjects:Computer Science > Historical Collection(Till Dec 2001)
ID Code:61
Deposited By:User autouser
Deposited On:08 November 2005