Title: Two-stage adaptive weight vector design method for decomposition based many-objective evolutionary algorithm
Authors: Xiaofang Guo; Yuping Wang; Xiaozhi Gao
Addresses: School of Science, Xi'an Technological University, Xi'an, Shaanxi, China ' School of Science, Xi'an Technological University, Xi'an, Shaanxi, China ' School of Computing, University of Eastern Finland, Kuopio, Finland
Abstract: In order to acquire the searching information of the irregular shape of the PF accurately, this paper proposes a two-stage adaptive weight vector design approach for decomposition based many objective evolutionary algorithms. Firstly, a preset initial weight vector generation strategy based on the crowding information is proposed to capture the effective area of the distribution of population. Next, the self-organised mapping (SOM) weight vector design method is adopted to generate weight vectors to capture the topological structure of the distribution of PF. In addition, a new composite aggregation function combined Lp metric distance with angle-distance is proposed to improve both convergence and diversity. The performance of our algorithm is examined using a total of 10 problems with both regular and irregular PF. The experimental results show that the proposed method can significantly improve the convergence and diversity performance with irregular shape of PF.
Keywords: many-objective evolutionary algorithm; two stage; crowding; self-organised mapping; SOM; Lp-metric.
International Journal of Ad Hoc and Ubiquitous Computing, 2022 Vol.39 No.1/2, pp.58 - 69
Received: 06 Oct 2020
Accepted: 24 Nov 2020
Published online: 18 Feb 2022 *