Title :
Running races with Fraser´s recombination
Author :
Fogel, David B. ; Fraser, Alex S.
Author_Institution :
Natural Selection Inc., La Jolla, CA, USA
Abstract :
Several recombination operators have been proposed in evolutionary computation. Standard procedures include one-point, two-point and uniform crossover. Little attention, however, has been given to a recombination operator that preceded each of these, which was offered by A.S. Fraser (1957). Fraser´s recombination assigns a variable probability for crossing over between two solutions at each locus. This operator subsumes the three standard forms of crossover. Experiments are conducted on a set of 10 test functions where individuals in a population are tagged to recombine in one of the four above-mentioned ways. Attention is focused on the relative performance of the various methods of recombination. Uniform crossover is seen to offer the greatest utility on the test functions, and Fraser´s recombination repeatedly outperforms one-point and two-point crossover. Several issues concerning the experimental design are discussed
Keywords :
design of experiments; evolutionary computation; mathematical operators; evolutionary computation; experimental design; one-point crossover; race running; recombination operators; relative performance; tagged individuals; test functions; two-point crossover; uniform crossover; variable probability; Biological system modeling; Biology computing; Books; Computational modeling; Couplings; Design for experiments; Evolutionary computation; Genetic algorithms; Performance evaluation; Testing;
Conference_Titel :
Evolutionary Computation, 2000. Proceedings of the 2000 Congress on
Conference_Location :
La Jolla, CA
Print_ISBN :
0-7803-6375-2
DOI :
10.1109/CEC.2000.870789