Generalised Pattern Search with Restarting Fitness Landscape Analysis

Abstract Fitness landscape analysis for optimisation is a technique that involves analysing black-box problems to extract pieces of information about the problem, which can beneficially inform design optimiser. Thus, algorithm aims address specific features detected during problem. Similarly, designer understand behaviour algorithm, even though problem unknown and performed via metaheuristic method. algorithmic made using fitness be seen as an example explainable AI in domain. The present paper proposes framework performs designs Pattern Search (PS) on basis results analysis. implemented restarting fashion: at each restart, refines updates pattern matrix used by PS. A computationally efficient implementation also presented this study. Numerical show proposed clearly outperforms standard PS another based Furthermore, two instances considered study are competitive with popular algorithms literature.