Journal
INFORMATION SCIENCES
Volume 422, Issue -, Pages 122-143Publisher
ELSEVIER SCIENCE INC
DOI: 10.1016/j.ins.2017.09.002
Keywords
Composite differential evolution; Multiple-elites-guided; Adaptive control parameters; Shift mechanism; Global numerical optimization
Categories
Funding
- National Natural Science Foundation of China [61772345, 61402291, 61402294, 61402534]
- Foundation for Distinguished Young Talents in Higher Education of Guangdong, China [2014KQNCX129]
- Major Fundamental Research Project in the Science and Technology Plan of Shenzhen [JCYJ20160310095523765, JCYJ20160307111232895, JCYJ20140418181958501, JCYJ20140828163633977]
Ask authors/readers for more resources
The performance of differential evolution (DE) has been significantly influenced by trial vector generation strategies and control parameters. Various powerful trial vector generation strategies with adaptive parameter adjustment methods such that the population generation is guided by the elites have been proposed. This paper aims to strengthen the performance of DE by compositing these powerful trial vector generation strategies, making it possible to obtain the guidance of each individual from multiple elites concurrently and independently. In this manner, the deleterious behavior in which an individual is misguided by various local optimal solutions into unpromising areas could be restrained to a certain extent. An adaptive multiple-elites-guided composite differential evolution algorithm with a shift mechanism (abbreviated as AMECoDEs) has been proposed in this paper. This algorithm concurrently employs two elites-guided trial vector generation strategies for each individual to generate two candidate solutions accordingly, and the best one is adopted to participate in the selection. Moreover, a hovel shift mechanism is established to handle stagnation and premature convergence issues. AMECoDEs has been tested on the CEC2014 benchmark functions. Experimental results show that AMECoDEs outperforms various classic state-of-the-art DE variants and is better than or at least comparable to various recently proposed DE methods. (C) 2017 Elsevier Inc. All rights reserved.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available