Accession Number : ADA302139

Title :   New Methods for Nonlinear Optimization.

Descriptive Note : Final rept. 15 May 91-30 Jun 94,

Corporate Author : COLORADO UNIV AT BOULDER DEPT OF COMPUTER SCIENCE

Personal Author(s) : Schnabel, Robert B. ; Byrd, Richard H.

PDF Url : ADA302139

Report Date : 30 AUG 1994

Pagination or Media Count : 11

Abstract : Over the course of this research grant, considerable progress was made in all the areas discussed in the proposal, namely limited memory methods for problems with constraints, tensor methods for large sparse nonlinear problems and for constrained optimization, and trust regions methods for nonlinearly constrained optimization. In addition, substantial progress was made in the development of large scale global optimization methods for molecular configuration problems, a topic supported in part by other agencies but one in which ARO has expressed considerable interest as well. We summarize the work in these areas in Sections 1-4. In addition, we have worked on several other topics, including symmetric-rank one update methods for unconstrained optimization, implementations of the linear algebraic operations of the BFGS method on sequential and parallel computers, and parallel methods for solving block bordered systems of nonlinear equations. We summarize this work very briefly in Section 5. Section 6 contains a listing of publications and reports supported by this grant, and Section 7 contains a list of research personnel supported by this grant.

Descriptors :   *MATHEMATICAL MODELS, *OPTIMIZATION, ALGORITHMS, SOFTWARE ENGINEERING, COMPUTATIONS, PARALLEL PROCESSING, MATHEMATICAL PROGRAMMING, NONLINEAR SYSTEMS, APPROXIMATION(MATHEMATICS), LEAST SQUARES METHOD, CONVERGENCE, NUMERICAL METHODS AND PROCEDURES, SPARSE MATRIX, APPLIED MATHEMATICS, NONLINEAR ALGEBRAIC EQUATIONS, LINEAR ALGEBRA.

Subject Categories : Operations Research

Distribution Statement : APPROVED FOR PUBLIC RELEASE