Primal and dual convergence of a proximal point exponential penalty method for linear programming
Abstract
We consider the diagonal inexact proximal point iteration uk - uk-1/?k ? - ? ? k f (uk, rk) + vk where f(x, r) = cTx + r ? exp[(Aix -bi)/r] is the exponential penalty approximation of the linear program min(cTx : Ax ? b}. We prove that under an appropriate choice of the sequences ?k, ?k and with some control on the residual vk, for every rk ? 0+ the sequence uk converges towards an optimal point u? of the linear program. We also study the convergence of the associated dual sequence ?ik = exp[(Aiuk - bi) / rk] towards a dual optimal solution.
Más información
Título según WOS: | Primal and dual convergence of a proximal point exponential penalty method for linear programming |
Título según SCOPUS: | Primal and dual convergence of a proximal point exponential penalty method for linear programming |
Título de la Revista: | MATHEMATICAL PROGRAMMING |
Volumen: | 93 |
Número: | 1 |
Editorial: | SPRINGER HEIDELBERG |
Fecha de publicación: | 2002 |
Página de inicio: | 87 |
Página final: | 96 |
Idioma: | English |
DOI: |
10.1007/s10107-002-0925-0 |
Notas: | ISI, SCOPUS |