Dual population-based incremental learning for problem optimization in dynamic environments

In recent years there is a growing interest in the research of evolutionary algorithms for dynamic optimization problems since real world problems are usually dynamic, which presents serious challenges to traditional evolutionary algorithms. In this paper, we investigate the application of Population-Based Incremental Learning (PBIL) algorithms, a class of evolutionary algorithms, for problem optimization under dynamic environments. Inspired by the complementarity mechanism in nature, we propose a Dual PBIL that operates on two probability vectors that are dual to each other with respect to the central point in the search space. Using a dynamic problem generating technique we generate a series of dynamic knapsack problems from a randomly generated stationary knapsack problem and carry out experimental study comparing the performance of investigated PBILs and one traditional genetic algorithm. Experimental results show that the introduction of dualism into PBIL improves its adaptability under dynamic environments, especially when the environment is subject to significant changes in the sense of genotype space.