Improved Variants of Gravitational Search Algorithm Based on 'Best-so-Far' Updating Mechanism

摘要

Gravitational Search Algorithm is a memory-less nature inspired optimization technique which keeps a record of the position of the "best- so- far" particle during the current as well as populations during previous iterations. The "best-so-far" particle or Lbest particle may not be the same as the best particle of the current population. With an objective to enhance the exploration of GSA, in this paper, three natural updating mechanisms of "best-so-far" particle are used to design three new variants of GSA, called IGSA-1, IGSA-2 and IGSA-3. The experiments are performed on a set of 23 benchmarks problems, divided into three category of problems. Based on the numerical analysis of results it is concluded that the overall performance of IGSA-1 is better than others on scalable unimodal function for 30 dimensional problems and low dimensional multimodal functions, whereas the performance of IGSA-3 is better than others on scalable multimodal functions having 30 dimension.