Modified proximal point algorithm for extended linear-quadratic programming (Q1203069)

This paper is concerned with a broad class of optimization problems known as extended linear-quadratic programming. The idea is to find a saddle point of a function in the extended linear quadratic form; the usual approach is via a proximal point algorithm (PPA). In this paper variations of the proximal point approach are proposed to deal with the case where the defining matrices in the extended linear-quadratic form are not both positive definite. Newton, Newton-like, and quasi-Newton variations are considered: results of a few numerical experiments are provided.