Adaptive Formation of Pareto Front in Evolutionary Multi-objective Optimization
Ö. Ciftcioglu, St Michael' · Evolutionary Computation · 2009
A novel adaptive approach for formation of the Pareto front in multi-objective optimization is presented. The approach is an adaptive stochastic search, where a relaxed dominance concept is introduced, and the relaxation angle is adapted during the search. This approach is a dual counterpart of gradient-based stochastic adaptive algorithms. In this duality the fitnessfunction is the dual of stochastic gradient, and the degree of relaxation is the dual of the step-size parameter. The adaptation is found to be significantly favorable for