Accession Number : ADA130903

Title :   Algorithms for Nonlinear Programming.

Descriptive Note : Final rept. 1 Aug 82-1 Jun 83,

Corporate Author : CITY COLL NEW YORK DEPT OF COMPUTER SCIENCE

Personal Author(s) : Goldfarb,Donald

PDF Url : ADA130903

Report Date : Jul 1983

Pagination or Media Count : 6

Abstract : Several algorithms in linear, quadratic, and nonlinear programming have been developed and analyzed. These include: (i) The development of relaxation methods for finding a feasible solution to a system of linear inequalities based upon generating surrgate constraints; (ii) The worst-case behaviour of the shadow-vertex simplex algorithm was shown to be the exponential; (iii) Necessary and sufficient conditions for versions of iterative methods, including the Jacobi, Gauss-Seidel and SOR methods, designed for solving equality constrained quadratic programs have been obtained; (iv) The development of numerically stable and efficient implementations of primal methods for quadratic programming; and (v) the development of optimal algorithms for estimating Jacobian and Hessian matrices arising in finite difference calculations. (Author)

Descriptors :   *Algorithms, *Nonlinear programming, Linear programming, Quadratic programming, Relaxation, Solutions(General), Inequalities, Exponential functions, Iterations, Stability, Efficiency, Optimization, Computations, Finite difference theory

Subject Categories : Theoretical Mathematics

Distribution Statement : APPROVED FOR PUBLIC RELEASE