A Synergistic Approach for Evolutionary Optimization

One of the major causes of premature convergence in Evolutionary Algorithm (EA) is loss of population diversity, which pushes the search space to a homogeneous or a near-homogeneous configuration. In particular, this can be a more complicated issue in case of high dimensional complex problem domains. In [13, 14], we presented two novel EA frameworks to curb premature convergence by maintaining constructive diversity in the population. The COMMUNITY_GA or COUNTER_NICHING_GA in [13] uses an informed exploration technique to maintain constructive diversity. In addition to this, the POPULATION_GA model in [14] balances exploration and exploitation using a hierarchical multi-population approach. The current research presents further investigation on the later model which synergistically uses an exploration controlling mechanism through informed genetic operators along with a multi-tier hierarchical dynamic population architecture, which allows initially less fit individuals a fair chance to survive and evolve. Simulations using a set of popular benchmark test functions showed promising results.