An improvement of database with local search mechanisms for genetic algorithms in large-scale computing environments

Recently, GA that uses large-scale computer systems comprised of massive processors has become feasible because of the emergence of super PC clusters and grid computation environments. Mechanisms to use massive computation resources laconically and to search effectively are necessary if large-scale computer systems are available. In this study, a new GA-specific database with the local search mechanism to assure the scalability of search performances against the number of computing resources is proposed. Our database possesses information of searched space, and in addition, the local search for nonsearched spaces is applied using individuals stored in the database. To embed our database in a GA enables us to comprehend the quantitative rate of a searched region during searches. Applying this local search, the searched space can be expanded linearly in accordance with the increase in computing resources and the exhaustive search is guaranteed under infinite computations. The features of the introduced GA are discussed with reference to several types of experiments. This method was applied to primitive functions and test functions of continuous optimization problems. Through such experiments, it was shown that our method ensures an effective exhaustive search and has almost the same performance as a conventional GA.