The combination algorithms of local search heuristics and genetic algorithms are often called "Genetic Local Search" (GLS). The GLS algorithms have been applied to combinatorial optimization problems, and its effectiveness has been recognized very well so far. Since the GLS algorithms contain many candidate individuals to find very good approximate solutions, it seems to be reasonable to suggest that rather than an use of fewer individuals. In general, a fixed number of many individuals has

more »

... set by making good use of researcher's experience or suggestions from the past in most cases. In this paper, we firstly investigate the effectiveness of the GLS algorithm, which contains only two individuals, by comparing with the GLS using many ones for benchmarks of the traveling salesman problem. It is shown that both GLS algorithms have the same condition except for only the number of individuals, and very good costs of solutions obtained by the algorithms are provided. From our experimental results, we demonstrate that the GLS using many individuals has good performance, however, it is not always suitable to use many individuals by secondly comparing with a novel 'genetic' approach that also contains only two solutions.