D. Mongru and C.W. Dawson (UK)
Genetic algorithms, linkage, inversion, introns.
In this paper we present a modification to the canonical Genetic Algorithm (CGA). The use of mixing and linkage can improve the GA search by helping to propagate highly fit sub-sections of solutions to the next generation. The inversion operator, despite being created with the original GA almost forty years ago, still has questionable worth. This paper introduces an algorithm which overcomes the problems of disruption associated with inversion by using introns to help restrict gene movement. Initial experiments show favourable results on a test-suite of problems, with results of a similar quality to those of the CGA, inversion and intron algorithms. There was an increase in quality as the problem difficulty increased. This indicates a scalability to the algorithm which could prove useful when applying the algorithm to real-world problems.
Important Links:
Go Back
