Summary of Research under ARO Contract DAAG-79-C-0023.
Abstract
This research has investigated several topics in solving unconstrained minimization, systems of nonlinear equations, and indefinite linearly constrained minimization problems. In unconstrained minimization, a new projected update was developed and tested. While marginal improvements over the BFGS were found, they probably do not justify the use of the new update. The use of conic models for unconstrained minimization problems when analytic or finite difference derivatives are available was also investigated, and a new algorithm was developed and tested. The results show reasonable improvements in many cases, and indicate that conic algorithms for minimization should continue to be considered. Recently we have also developed a tensor algorithm for solving systems of nonlinear equations, and the preliminary computational results show considerable improvements over the corresponding standard algorithm, especially on problems where the Jacobian at the solution is singular. This approach seems to hold excellent promise. Other research in linearly constrained minimization and deriving and analyzing least change secant updates also is reported. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 11, 1981
- Accession Number
- ADA105951
Entities
People
- Robert B. Schnabel
Organizations
- University of Colorado Boulder