New class of multiplicative algorithms for solving of entropy-linear programs

被引:6
|
作者
Popkov, Yu. S. [1 ]
机构
[1] Inst Syst Anal, Moscow 117312, Russia
关键词
entropy maximization; multiplicative algorithms; row-action procedures; active variables; feedback control; G-convergence; exponential Lagrange multipliers;
D O I
10.1016/j.ejor.2005.01.069
中图分类号
C93 [管理学];
学科分类号
12 ; 1201 ; 1202 ; 120202 ;
摘要
The general entropy-linear program (ELP) is considered. Two types of the new coordinate-wise multiplicative algorithms with p-active variables and feedback control with respect to dual variables and mixed type (dual and primal variables) are proposed for solving the problem. Study of algorithms convergence is based on a stability analysis of the auxiliary differential equations that are continuous analogues of the algorithms. (c) 2005 Elsevier B.V. All rights reserved.
引用
收藏
页码:1368 / 1379
页数:12
相关论文
共 50 条