An Imperialistic Strategy Approach to Continuous Global Optimization Problem

The paper is introducing the principles of a new global optimization strategy, Imperialistic Strategy (IS), applied to the Continuous Global Optimization Problem (CGOP). Inspired from existing multi-population strategies, like the Island Model (IM) approaches to parallel Evolutionary Algorithms (EA) and the Imperialistic Competitive Algorithm (ICA), the proposed IS method is considered an optimization strategy for the reason that it can integrate other well-known optimization methods, which in the context are regarded as sub-methods (although in other contexts they are prominent global optimization methods). Four optimization methods were implemented and tested in the roles of sub-methods: Genetic Algorithm (GA) (a floating-point representation variant), Differential Evolution (DE), Quantum Particle Swarm Optimization (QPSO) and Artificial Bee Colony (ABC). The optimization performances of the proposed optimization methods were compared on a test bed of 9 known multimodal optimization problems by applying an appropriate testing methodology. The obtained increased success rates of IS multi-population variants compared to the success rates of the optimization sub-methods run separately, combined with the increased computing efficiencies possible to be perceived for parallel and distributed implementations, demonstrated that IS is a promising approach to CGOP.