New memetic self-adaptive firefly algorithm for continuous optimisation

作     者：Galvez, Akemi Iglesias, Andres 

作者机构：Univ Cantabria ETSI Caminos Canales & Puertos Dept Appl Math & Computat Sci Avda Castros S-N Santander 39005 Spain Toho Univ Fac Sci Dept Informat Sci 2-2-1 Miyama Funabashi Chiba 2748510 Japan 

出 版 物：《INTERNATIONAL JOURNAL OF BIO-INSPIRED COMPUTATION》 (国际生物启发计算杂志)

年 卷 期：2016年第8卷第5期

页      面：300-317页

核心收录：

学科分类：0710[理学-生物学] 07[理学] 09[农学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：Computer Science National Programme of the Spanish Ministry of Economy and Competitiveness [TIN2012-30768] Toho University (Funabashi, Japan) University of Cantabria (Santander, Spain) 

主　　题：firefly algorithm FFA continuous optimisation nature-inspired metaheuristic memetic approach breakpoint location problem 

摘      要：The firefly algorithm is a recent nature-inspired algorithm that is receiving increasing attention from the scientific community during the last few years. One of its most promising variants is given by the memetic self-adaptive firefly algorithm (MSA-FFA), recently introduced to solve combinatorial problems. In this paper we propose a modification of the original MSA-FFA for continuous optimisation problems. The most important features of our method are: the problem-dependent selection of control parameters for self-adaptation, a simple population model providing an adequate trade-off between exploration and exploitation, and the use of an adaptive-size Luus-Jaakola random local search. This new method is applied to solve a very difficult real-world continuous optimisation problem arising in geometric modelling and manufacturing. The paper also provides the first reliable, standardised benchmark for this optimisation problem. This benchmark is used for a comparative analysis of our method with respect to some of the most popular nature-inspired algorithms. Our results show that the proposed method outperforms previous approaches (including the standard firefly algorithm) for most of the instances in the benchmark.