Asynchronous Island Parallel GA Using Multiform Subpopulations

Island Parallel GA divides a population into subpopulations and assigns them to processing elements on a parallel computer. Then each subpopulation searches the optimal solution independently, and exchanges individuals periodically. This exchange operation is called migration. In this research, we propose a new algorithm that migrants are exchanged asynchronously among multiform subpopulations which have di erent search conditions. The e ect of our algorithm on combinational optimization problems was veri ed by applying the algorithm to Knapsack Problem and Royal Road Functions using parallel computer CRAY-T3E. We obtained the results that our algorithm maintained the population's diversity e ectively and searches building blocks e ciently.