Survival of the sickest: a site-specific recombination operator for accelerated function optimization

This paper describes experiments with a new crossover operator which is based on the mechanism of biological site-specific recombination. By using hill-climbing to gather additional information about the fitness landscape, it increases the constructive power of crossover. The nature of the operator calls for a somewhat unusual selection strategy which, in contrast with traditional methods, automatically selects a relatively unfit member of the population to be a parent. When applied to difficult continuous-variable function optimization problems, the operator is seen to perform better than standard one-point crossover in terms of quality of solutions found and its speed in finding them (as measured by the number of function evaluations carried out).