Genetic algorithms + data structures = evolution programs (3rd ed.)
Genetic algorithms + data structures = evolution programs (3rd ed.)
| Hi-index | 0.00 |
A typical genetic algorithm uses a constant mutation rate or reduces the mutation rate over the generation. Generally, the degree of perturbation by crossover operators gets weaker as the generations go by. Hybrid GAs, which use a local optimization heuristic, strongly drive the offspring to a chromosome similar to or the same as one of the parents. We suspect that one needs to raise the degree of mutation in the late stages of a hybrid GA, contrary to the practice. The experimental results supported our suspection, by showing performance improvement over two philosophically representative mutations: the traditional fixed-rate mutation and the non-uniform mutation. We used two representative NP-hard problems in the experiments: the graph bisection problem and the traveling salesman problem.