Greediness Characterization of Genetic Algorithms

This research investigates the effect of selection, replacement, and generational gap on the selective pressure, indicated by the greediness of the genetic algorithms. An a priori ordering of the greediness of genetic algorithms is constructed based on the nature of selection strategy that favors fitter candidate solutions. For the implementers of genetic algorithms, this ordering represents the design space of genetic algorithms involving selection, replacement, and generation gap on which design decision should be taken. These relative orderings of the greediness were examined by experimental runs on timetabling benchmark problems. The results suggest that for a fixed computational budget, a high greediness can be achieved by biased selection with a large generational gap and elitism.