摘要：Migration o f individuals allows a fruitful i nteraction b etween subpopulations in the islandmodel, a well known d istributed approach for evolutionary computing, where separatesubpopulations evolve in p arallel. This model i s well suited for a distributed environmentrunning a Single Program Multiple Data (SPMD) scheme. Here, the same Genetic Algorithm(GA) is replicated in many processors and attempting better convergence, through an expectedimprovement on genetic diversity, selected individuals are e xchanged p eriodically. Forexchanging, an individual is selected from a source subpopulation and then exported towards atarget subpopulation. Usually, the imported string is accepted on arrival and then inserted intothe target subpopulation. Our earlier experiments on controlled migration showed animprovement on results when contrasted against t hose obtained b y conventional migrationapproaches.This paper describes extended implementations of alternative strategies to oversee migration inasynchronous schemes for an island model and enlarges a previous work on three processorswith a set of softer testing functions [9]. All of them try to d ecrease the risk of prematureconvergence. A first strategy attempts to p revent unbalanced p ropagation o f genotypes byapplying an acceptance threshold p arameter to each incoming string. A second one permitsindependent evolution of subpopulations and acts only when a possible stagnation is detected.In such condition an attempt to evade falling towards a local optimum is done by inserting anexpected d issimilar individual t o improve genetic diversity. A third alternative strategycombines both previous mentioned strategies. The results presented are those obtained on thefunctions that showed to be more difficult for the island model using a replication of a simpleGA. A description o f the c orresponding system architecture supporting the PGAimplementation is described and results for the parallel distributed approach among 3, 6 and 12processors is discussed
