Landscape-based adaptive operator selection mechanism for differential evolution

Karam M. Sallam, Saber M. Elsayed, Ruhul A. Sarker, Daryl L. Essam

Research output: Contribution to journalArticlepeer-review

23 Citations (Scopus)

Abstract

Over the last two decades, many different differential evolution algorithms for solving optimization problems have been introduced. Although most of these algorithms used a single mutation strategy, several with multiple mutation strategies have recently been proposed. Multiple-operator-based algorithms have been proven to be more effective and efficient than single-operator-based algorithms for solving a wide range of benchmark and practical problems. In these algorithms, adaptive operator selection mechanisms are generally applied to place greater emphasis on the best-performing evolutionary operators based on their performance histories for generating new offspring. In this paper, we investigate using problem landscape information in an adaptive operator selection mechanism. For this purpose, a new algorithm, which considers both this problem landscape information and the performance histories of the operators, for dynamically selecting the most suitable differential evolution operator during the evolutionary process, is proposed. The contributions of each component of the selection mechanism are analyzed and the performance of the proposed algorithm is evaluated by solving 45 unconstrained optimization problems. The results demonstrate the effectiveness and superiority of the proposed algorithm to state-of-the-art algorithms.

Original languageEnglish
Pages (from-to)383-404
Number of pages22
JournalInformation Sciences
Volume418-419
DOIs
Publication statusPublished - 1 Dec 2017
Externally publishedYes

Fingerprint

Dive into the research topics of 'Landscape-based adaptive operator selection mechanism for differential evolution'. Together they form a unique fingerprint.

Cite this