An Effective Line Search for the Subgradient Method |
| |
| Authors: | C Beltran F J Heredia |
| |
| Institution: | (1) Researcher, Logilab, HEC, University of Geneva, Geneva, Switzerland;(2) Professor, Department of Statistics and Operations Research, Polytechnical University of Catalonia, Barcelona, Spain |
| |
| Abstract: | One of the main drawbacks of the subgradient method is the tuning process to determine the sequence of steplengths. In this paper, the radar subgradient method, a heuristic method designed to compute a tuning-free subgradient steplength, is geometrically motivated and algebraically deduced. The unit commitment problem, which arises in the electrical engineering field, is used to compare the performance of the subgradient method with the new radar subgradient method.Communicated by M. SimaanThis research was supported by the Spanish Government, CICYT Grant TAP99-1075-C02-01. We acknowledge the technical support from Logilab (HEC, University of Geneva) and especially the valuable remarks and suggestions of the referees. |
| |
| Keywords: | Lagrangian relaxation multiplier updating subgradient method radar unit commitment |
| 本文献已被 SpringerLink 等数据库收录！ |
|
