Line-breeding is an interesting mating strategy adapted to genetic algorithms. Most experiments on line-breeding have used multimodal functions as a testbed. In this work, we focus on combinatorial optimization problems. We chose the multiple constrained knapsack problem as our testbed. Several line-breeding schemes are explored, from the naive version to more advanced versions with sharing and niching. A new mechanism, dynamic mutation, is also proposed. Under this mechanism, if two individuals are very close to each other, instead of mating, a self-fertilization (i.e., a special mutation) is applied. The experiments presented in this paper show that a line-breeding scheme which uses multiple distanced champions can achieve the best performance. The paper also shows that dynamic mutation can improve not only the quality of solutions but can also identify more local optima.
CITATION STYLE
Yang, R. (1998). Line-breeding schemes for combinatorial optimization. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 1498 LNCS, pp. 448–457). Springer Verlag. https://doi.org/10.1007/bfb0056887
Mendeley helps you to discover research relevant for your work.