Examining algorithm behavior using recurrence quantification and landscape analyses
Mario Andrés Muñoz · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2022
Differences in performance between algorithms can be attributed to the interaction between their unique rule-sets and the characteristics of the instance's landscape. However, understanding this interaction can be difficult because algorithms are often composed of multiple elements, and in the worst cases are described using opaque notation and metaphors. In this paper, we introduce a methodology for the behavioral analysis of optimization algorithms, based on comparing algorithm dynamics in a given problem instance. At the methodology's core lays the hypothesis that if two algorithms, with the exact same initial conditions, have similar dynamics, then their rule-sets are also similar. An examination of Grey Wolf Optimization, shows that it exhibits bias leading to similar behavioral patterns regardless of the function.