An Empirical Comparison of Selection Methods in Evolutionary Algorithms

摘要: Selection methods in Evolutionary Algorithms, including Genetic Evolution Strategies (ES) and Programming, (EP) are compared by observing the rate of convergence on three idealised problems. The first considers selection only, second introduces mutation as a source variation, third also adds evaluation noise. Fitness proportionate suffers from scaling problems: number techniques to reduce these illustrated. sampling errors caused roulette wheel tournament demonstrated. EP model is shown be equivalent an ES one form, surprisingly similar fitness another. Generational models remarkably immune noise, that retain parents much less so.