Enhancing meta-heuristic algorithms using center-based sampling at population level

Abstract

In recent years, center-based sampling has demonstrated impressive results to enhance the efficiency and effectiveness of meta-heuristic algorithms. The strategy of center-based sampling can be utilized at either the operation and/or population levels. Despite the overall efficiency of the center-based sampling in population-based algorithms, utilization at the operation level requires customizing the strategy for a specific algorithm which degrades the scheme’s generalization. This study proposes a center-based sampling at the population level, which is operation-independent and correspondingly can be embedded in any population-based optimization algorithm. In classic mutation and crossover operators, the number of parents involved is a few, causing ineffective exploration; however, the current proposed center-based sampling uses a multi-parent approach, which results in multiple center-based solutions. In this thesis, two proposed schemes, namely, 1) Clustering center-based sampling and 2) Average ranking center-based sampling, are applied to enhance population-based single- and multi-objective optimization algorithms, respectively, in order to enhance their exploration and exploitation capabilities. The conducted comprehensive center-based experiments are a novel strategy to enhance population based mechanistic algorithms. In order to assess the performance of proposed schemes, the proposed strategy is applied to single- and multi-objective optimization problems and experimented with CEC-2017 benchmark functions. The experimental outcomes confirm that the proposed clustering center-based and ranking center-based samplings have a crucial positive impact on convergence rate of various families of optimization algorithms.