Productive fitness in diversity-aware evolutionary algorithms

Abstract In evolutionary algorithms, the notion of diversity has been adopted from biology and is used to describe distribution a population solution candidates. While it known that maintaining reasonable amount often benefits overall result optimization process by adjusting exploration/exploitation trade-off, little about what optimal. We introduce productive fitness based on effect specific candidate some generations down path. derive final fitness, which ideal target for any process. Although inefficient compute, we show empirically allows an posteriori analysis how well given hit providing insight into why diversity-aware performs better.