Regresar

A study on interval to swtich combination of objectives considered in pareto partial dominance MOEA

Abstract:

Pareto partial dominance MOEA (PPD-MOEA) ranks individuals in the population by using only r (<m) objectives selected from m objective functions, which makes easier to discriminate solutions than conventional Pareto dominance MOEAs even in many-objective optimization problems (MaOP). In PPD-MOEA, r objective functions are temporally switched among <inf>m</inf>C <inf>r</inf> combinations of objectives every I<inf>g</inf> generations to optimize all of the objective functions throughout the entire evolution process. To induce maximum search performance of PPD-MOEA, in this work, we study on appropriate setting for the interval parameter I<inf>g</inf> that switches combination of objectives. We analyze the search performance of PPD-MOEA by focusing the exhaustivity of <inf>m</inf>C<inf>r</inf> kinds of combination in the entire evolution process. Simulation results show that maximum search performance of PPD-MOEA using the optimal r* is achieved when we set the boundary interval I<inf>g</inf><sup>b</sup>, i.e., the exhaustivity is 1.0, in which all the <inf>m</inf>C<inf>r</inf>, combinations are once utilized in the entire evolution process of PPD-MOEA. © 2010 TSI Press.