Enhancing Brain Storm Optimization Through Optimum-Path Forest

Abstract

Among the many interesting meta-heuristic optimization algorithms, one can find those inspired by both the swarm and social behavior of human beings. The Brain Storm Optimization (BSO) is motivated by the brainstorming process performed by human beings to find solutions and solve problems. Such process involves clustering the possible solutions, which can be sensitive to the number of groupings and the clustering technique itself. This work proposes a modification in the BSO working mechanism using the Optimum-Path Forest (OPF) algorithm, which does not require the knowledge about the number of clusters beforehand. Such skill is pretty much relevant when this information is unknown and must be set. The proposed approach is evaluated in a set of six benchmarking functions and showed promising results, outperforming the traditional BSO and a second variant makes use of the well-known Self-Organizing Maps clustering technique.