Region encoding helps evolutionary computation evolve faster: a new solution encoding scheme in particle swarm for large-scale optimization

Jian, Jun-Rong ORCID: 0000-0003-1277-4516, Chen, Zong-Gan ORCID: 0000-0001-7585-5212, Zhan, Zhi-Hui ORCID: 0000-0003-0862-0514 and Zhang, Jun ORCID: 0000-0001-7835-9871 (2021) Region encoding helps evolutionary computation evolve faster: a new solution encoding scheme in particle swarm for large-scale optimization. IEEE Transactions on Evolutionary Computation, 25 (4). pp. 779-793. ISSN 1089-778X

Abstract

In the last decade, many evolutionary computation (EC) algorithms with diversity enhancement have been proposed to solve large-scale optimization problems in big data era. Among them, the social learning particle swarm optimization (SLPSO) has shown good performance. However, as SLPSO uses different guidance information for different particles to maintain the diversity, it often results in slow convergence speed. Therefore, this article proposes a new region encoding scheme (RES) to extend the solution representation from a single point to a region, which can help EC algorithms evolve faster. The RES is generic for EC algorithms and is applied to SLPSO. Based on RES, a novel adaptive region search (ARS) is designed to on the one hand keep the diversity of SLPSO and on the other hand accelerate the convergence speed, forming the SLPSO with ARS (SLPSO-ARS). In SLPSO-ARS, each particle is encoded as a region so that some of the best (e.g., the top ${P}$ ) particles can carry out region search to search for better solutions near their current positions. The ARS strategy offers the particle a greater chance to discover the nearby optimal solutions and helps to accelerate the convergence speed of the whole population. Moreover, the region radius is adaptively controlled based on the search information. Comprehensive experiments on all the problems in both IEEE Congress on Evolutionary Computation 2010 (CEC 2010) and 2013 (CEC 2013) competitions are conducted to validate the effectiveness and efficiency of SLPSO-ARS and to investigate its important parameters and components. The experimental results show that SLPSO-ARS can achieve generally better performance than the compared algorithms.

Dimensions Badge

Altmetric Badge

Item type Article
URI https://vuir.vu.edu.au/id/eprint/45255
DOI 10.1109/TEVC.2021.3065659
Official URL https://ieeexplore.ieee.org/document/9377474
Subjects Current > FOR (2020) Classification > 4602 Artificial intelligence
Current > Division/Research > Institute for Sustainable Industries and Liveable Cities
Keywords artificial intelligence, evolutionary computation, algorithms, large scale optimisation
Citations in Scopus 40 - View on Scopus
Download/View statistics View download statistics for this item

Search Google Scholar

Repository staff login