Distributed population based incremental learning
Rui Rodrigues · 2005
Abstract- Probabilistic model building genetic algorithms (PMBGA) are powerful search techniques that are used suc-cessfully to solve hard computational problems. We exploit the natural bridge island concept in the Parallel PMBGA context. Bridge Island is an island that all the best elements arriving from evolving islands have to pass before its migra-tion to other islands occurs. Some ways of best elements selection to migrate and also some ways of merging informa-tion between the migrating element and the receiver element will be tested. The main objective is to compare the rela-tive performance between distributed PBIL and isolated PBIL. The parallel models that will be used are the master-slave topology, island model and coarse-grained.