A Comparison of Single-Based Versus Population-Based Search Algorithms in the Optimization of Fuzzy Systems

In this article, the performance of two types of search algorithm approaches in the optimization of fuzzy systems are studied and analyzed. We compared the Generalized Pattern Search and the Simulated Annealing Algorithm, classified as single-based algorithms, against the Genetic Algorithm and Particle Swarm Optimization, classified as population-based algorithms. Algorithms classified as single-based algorithms use a candidate solution that improves through an iterative process, while algorithms classified as population-based algorithms base their search on a set of solutions that interact collaboratively. The performance of the algorithms will be measured when optimizing Mamdani-type fuzzy systems, with a smooth output surface, taking as a metric the number of evaluations carried out by the algorithms. Finally, the convergence of the algorithms will be analyzed.