We propose two entropy-based diversity measures for evaluating population diversity in a genetic algorithm (GA) applied to the traveling salesman problem (TSP). In contrast to a commonly used entropy-based diversity measure, the proposed ones take into account high-order dependencies between the elements of individuals in the population. More precisely, the proposed ones capture dependencies in the sequences of up to
+ 1 vertices included in the population (tours), whereas the commonly used one is the special case of the proposed ones with
= 1. We demonstrate that the proposed entropy-based diversity measures with appropriate values of
evaluate population diversity more appropriately than does the commonly used one.
Bitte loggen Sie sich ein, um Zugang zu diesem Inhalt zu erhalten