A Search Cost Reduction in Particle Swarm Optimizers Using a Blindness Factor

Abstract Many algorithmic procedures are available in the literature able to face optimization problems. Competent deterministic approaches deal efficiently with convex objective functions where the practitioner bounds the search spaces. However, the objective functions of real-world optimization problems are seldom convex. They can be multi-modal, non-convex, non-separable, non-symmetric, and several fitness surfaces have many local optima far from the global optimum (deceptive problems), making them even more challenging to find acceptable solutions with fewer objective function calls throughout the search process. With this in mind, this study aims to propose a new feature for non-deterministic approaches, such as the particle swarm optimization (PSO) algorithm, that can reduce the number of objective function calls and still find at least a suitable solution in the context of the problem. This paper introduces the \textit{blindness factor}, an integer parameter that allows a particle to consecutively move through the search space knowing nothing about its quality, i.e., this factor controls the total number of function queries a particle can do and when the function is consulted. Computational experiments showed the effectiveness of the blindness factor in the PSO algorithm on different benchmark optimization problems.