On the Solution of Indefinite Systems Arising in Nonlinear Optimization

Silvia Bonettini, Federica Tinti, Valeria Ruggiero

Abstract


This work is concerned with the solution of a class of symmetric
indefinite linear systems of equations, typically arising in
nonlinear optimization: indeed, the solution of such system is
crucial for determining the search direction of many
Interior--Point methods. Our approach is based on the preconditioned
conjugate gradient method with the
choice of a quasidefinite preconditioner and of a suitable
Cholesky--like factorization subroutine. We show a numerical comparison of the
performances of the preconditioned conjugate gradient method
applied to different formulations of the linear system.

[DOI: 10.1685 / CSC06025] About DOI

Full Text:



[DOI: 10.1685/] About DOI

Url Resolver: : http://dx.doi.org/10.1685/





Creative Commons License   Except where otherwise noted, content on this site is
  licensed under a Creative Commons 2.5 Italy License