A new balanced direction for multi-objective evolutionary algorithms

  • Nguyen Long
  • Nguyen Xuan Hung
  • Nguyen Thi Hien
  • Bui Thu Lam

Abstract

This paper suggests to use a new improvement direction for multi-objective evolutionaryalgorithms. In DMEA-II, two improvement directions(convergence and spread) are used for the guidanceduring evolutionary processes. Based on those directions and the balance between exploration andexploitation, we determined a new improvement direction to keep DMEA-II to be better on the balanceof convergence and diversity.To validate the performance of the new improvedversion of DMEA-II, we carried out a case studyon several test problems and comparison with wellknown MOEAs, it obtained quite good results onprimary performance metrics, namely the generationdistance, inverse generation distance and hypervolume. Our analysis on the results indicates that,the usage of proposed direction may make DMEAII to be improved in keeping balanced betweenconvergence and diversity at each generation duringthe search
Published
2016-11-18
Section
Regular Articles