Genetic Algorithms (GAs) are efficient non-gradient stochastic search methods and Parallel GAs (PGAs) are proposed to overcome the deficiencies of the sequential GAs, such as low speed, aptness to local convergence, e...Genetic Algorithms (GAs) are efficient non-gradient stochastic search methods and Parallel GAs (PGAs) are proposed to overcome the deficiencies of the sequential GAs, such as low speed, aptness to local convergence, etc. However, the tremendous increase in the communication costs accompanied with the parallelization stunts the further improvements of PGAs. This letter takes the decrease of the communication costs as the key to this problem and advances a new Migration Scheme based on Schema Theorem (MSST). MSST distills schemata from the populations and then proportionately disseminates them to other populations, which decreases the total communication cost among the populations and arms the multiple-population model with higher speed and better scalability.展开更多
基金National Natural Science Foundation of China (No.60073012)National Science Foundation of Jiangsu, China(BK2001004)Visiting Scholar Foundation of Key Lab. in the University
文摘Genetic Algorithms (GAs) are efficient non-gradient stochastic search methods and Parallel GAs (PGAs) are proposed to overcome the deficiencies of the sequential GAs, such as low speed, aptness to local convergence, etc. However, the tremendous increase in the communication costs accompanied with the parallelization stunts the further improvements of PGAs. This letter takes the decrease of the communication costs as the key to this problem and advances a new Migration Scheme based on Schema Theorem (MSST). MSST distills schemata from the populations and then proportionately disseminates them to other populations, which decreases the total communication cost among the populations and arms the multiple-population model with higher speed and better scalability.
