A feasible-ratio control technique for constrained optimization

In constrained optimization problems (COPs), a crucial issue is that most constraint-handling evolutionary algorithms (EAs) approach the optimum either mainly from feasible regions or mainly from infeasible regions. This may result in bias in search of feasible and infeasible solutions. To address t...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:Information sciences Ročník 502; s. 201 - 217
Hlavní autori: Jiao, Ruwang, Zeng, Sanyou, Li, Changhe
Médium: Journal Article
Jazyk:English
Vydavateľské údaje: Elsevier Inc 01.10.2019
Predmet:
Constrained multi-objective optimization
Constrained optimization
Constraint-handling
Evolutionary computation
Constrained multi-objective optimization
Evolutionary computation
Constraint-handling
Constrained optimization
ISSN:0020-0255, 1872-6291
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Popis
Shrnutí:In constrained optimization problems (COPs), a crucial issue is that most constraint-handling evolutionary algorithms (EAs) approach the optimum either mainly from feasible regions or mainly from infeasible regions. This may result in bias in search of feasible and infeasible solutions. To address this issue, we propose a feasible-ratio control technique which controls the ratio of feasible solutions in the population. By using the control technique, an EA can maintain the search balance from feasible and infeasible regions. Based on this technique, we propose a constraint-handling EA, named FRC-CEA. It consists of two-stage optimization. In the first stage, an enhanced dynamic multi-objective evolutionary algorithm (DCMOEA) with the feasible-ratio control technique is adopted to handle constraints. In the second stage, a commonly used differential evolution (DE) is used to speed up the convergence. The performance of the proposed method is evaluated and compared with six state-of-the-art constraint-handling algorithms on two sets of benchmark test suites. Experimental results suggest that the proposed method outperforms or is highly competitive against the compared algorithms on most test problems.
ISSN:0020-0255
1872-6291
DOI:10.1016/j.ins.2019.06.030