Traditional approaches to real-valued function optimization using evolutionary computational methods tend to use either self-adaptive operators (as in the case of evolutionary programming), or population-based operators (as in the case of most real-valued genetic algorithms). However, in general, most population-based operators are limited in scope to the use of at most two or three parent individuals. In this paper we explore an alternative populationbased form of adaptation for evolutionary computation, Guided Gaussian Mutation (GGM), which is designed specifically as a localized search operator. This operator is the first of a larger class of Cohort Driven Operators (CDOs) which we define here. Experimental results using GGM in a standard genetic algorithm framework on a series of test problems show impressive improvement over standard evolutionary programming.
展开▼
