A unifying approach to entropy-like multiplier methods in convex programming

A unifying approach to entropy-like multiplier methods in convex programming

0.00 Avg rating0 Votes
Article ID: iaor2008966
Country: Hungary
Volume: 22
Issue: 1
Start Page Number: 63
End Page Number: 83
Publication Date: Jan 2005
Journal: Alkalmazott Matematikai Lapok
Authors:
Abstract:

Entropy-like multiplier methods solve the convex constrained minimization problem by solving its dual. To regularize optimization process a distance function parametrized by the current estimate of the Lagrangian multipliers is added to the Lagrangian of the problem, to enforce positivity of these multipliers the chosen function is a so-called entropy-like function which is defined on the positive orthant. In this paper we are concerned with externing available convergence results without giving some specific form to this entropy-like function. Several examples of applicable entropy-like functions are presented.

Reviews

Required fields are marked *. Your email address will not be published.