Abstract: | The ability to obtain multiple distinct solutions in a single run is an important, though often forgotten, practical advantage of genetic algorithms, and therefore methods that enhance this ability are desirable. The approach taken here is inspired by nature: sexually reproducing organisms do not mate indiscriminately - the choice of mate has a large impact upon the fitness of the organism's offspring, and by balancing exploration and exploitation, mate (or sexual) selection can lead to speciation effects which may enhance the genetic algorithm's ability to locate multiple distinct solutions. The investigation described here shows that sexual selection was able to enhance the genetic algorithm's ability to locate and maintain multiple distinct solutions, although it can interact detrimentally with other techniques designed for the same purpose.
|