The coordination of parallel search with common components

Abstract

The preservation of common components has been recently isolated as a beneficial feature of genetic algorithms. One interpretation of this benefit is that the preservation of common components can direct the search process to focus on the most promising parts of the search space. If this advantage can be transferred from genetic algorithms, it may be possible to improve the overall effectiveness of other heuristic search techniques. To identify common components, multiple solutions are required - like those available from a set of parallel searches. Results with simulated annealing and the Traveling Salesman Problem show that the sharing of common components can be an effective method to coordinate parallel search. © Springer-Verlag Berlin Heidelberg 2005.