Effects of External Archives on the Performance of Multi-Objective Evolutionary Algorithms on Real-World Problems

External archives have attracted more and more attention in the evolutionary multi-objective optimization (EMO) community. This is because a solution set selected from an external archive is usually better than the final population of an EMO algorithm. Whereas the effects of subset selection from external archives have already been investigated on artificial test problems, its effects on real-world problems have not been examined. In this paper, we examine the effects of subset selection from external archives for ten EMO algorithms on two real-world problem suites. Experimental results show that the performance improvement by subset selection is large for most algorithms and many problems but small for a few algorithms and a few problems (i.e., algorithm dependent and problem dependent).