A unified race algorithm for offline parameter tuning

摘要: This paper proposes uRace, a unified race algorithm for efficient offline parameter tuning of deterministic algorithms. We build on the similarity between stochastic simulation environment and algorithms, where element in latter is unknown problem instance given to algorithm. Inspired by techniques from optimization literature, uRace enforces fair comparisons among configurations evaluating their performance same training instances. It relies rapid statistical elimination inferior an increasingly localized search space quickly identify good settings. empirically evaluate applying it parameterized algorithmic framework loading problems at ORTEC, global provider software solutions complex decision-making problems, obtain competitive results set practical instances one world's largest multinationals consumer packaged goods.