Incorporating Neighbourhood Search Operators Into Genetic Algorithms
In this paper an abstract genetic algorithm (GA) is proposed which effectively merges local hill-climbing with recombination and population based selection in a general manner. This extension is possible because the traditional crossover can be resolved into two functions: one function is as a particular class of operator, which is actually distinct from the other which is the recombination itself. Thus traditional GAs can be classified as a special case of a more general approach in which recombination is applied along with other operators. In the work reported here, using the framework of an abstract GA, the performance of several operators as well as the effects of recombination are studied in the context of the graph bipartitioning problem.
KeywordsGenetic Algorithm Search Space Local Search Graph Transformation Hill Climber
Unable to display preview. Download preview PDF.
- Reeves, C.R. (1994) Genetic Algorithms and Neighbourhood Search. Proc. of AISB Workshop on Evolutionary Computing, Springer Verlag, Berlin, 115–130.Google Scholar
- Culberson, J.C. (1993) Mutation-crossover isomorphisms and the construction of discriminating functions. Evolutionary Computation (to appear)Google Scholar
- Hoehn, C. (1995) Embedding local search operators in a genetic algorithm for graph bipartitioning. In preparation.Google Scholar
- Hoehn, C. (1994) Heuristic Neighbourhood Search Operators for Graph Partitioning Tasks Proc. 10th International Conference on Systems Engineering, Coventry, UK, 469–476Google Scholar