Hybrid Multiobjective Differential Evolution Based on Positions of Individuals in Multiobjective Optimization

This paper proposes a hybrid multiobjective differential evolution (HMODE) framework based on positions of individuals in Pareto frontier (PF) to solve multiobjective optimization problem. Firstly, the hybrid multiobjective evolutionary algorithm (HMOEA) in HMODE is designed as the global search str...

Full description

Saved in:
Bibliographic Details
Main Authors: Zhang, Wenqiang, Yang, Diji, Wang, Yu, Qian, Zhan, Xu, Heyang, Gen, Mitsuo
Format: Conference Proceeding
Language:English
Subjects:
Convergence
differential evolution
Evolutionary computation
Heuristic algorithms
hybridization
multiobjective
Optimization
Silicon
Sociology
Statistics
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This paper proposes a hybrid multiobjective differential evolution (HMODE) framework based on positions of individuals in Pareto frontier (PF) to solve multiobjective optimization problem. Firstly, the hybrid multiobjective evolutionary algorithm (HMOEA) in HMODE is designed as the global search strategy to explore the entire solution space. HMOEA uses Pareto dominating and dominated relationship-based fitness function (PDDR-FF) to distinguish the nondominated and dominated individuals. The location information of individual can be determined by its PDDR-FF value. The elitist maintenance strategy based on PDDR-FF and simple selection strategy guarantee the capability of converging to the multiple directions of PF quickly and distributing along the PF uniformly. Secondly, differential evolution (DE) is combined with HMOEA as the local search to enhance the convergence and distribution performances on the elite population derived from HMOEA. Different PDDR-FF values of individuals clearly reflect their dominance relationships and position information in the PF. DE uses the position information of alternative individuals to finely tune the search directions to multiple areas of PF. Numerical comparisons indicate that the efficacy of HMODE outperforms HMOEA in convergence and distribution performances.
ISSN:2577-1655
DOI:10.1109/SMC.2018.00620