Statement of the Tuning Problem

Abstract

Metaheuristics are general algorithmic templates whose components need to be instantiated and properly tuned in order to yield a fully functioning algorithm. We call a configuration of a metaheuristic any possible instantiation of this template while with the expression tuning problem we denote the problem of properly instantiating the template, that is, the problem of choosing among the set of possible components and assigning specific values to all free parameters.

As practitioners know, metaheuristics are in general quite sensitive to the value of their parameters and a careful fine-tuning typically improves the performance in a significant way. This chapter is devoted to the definition of a formal framework in which the tuning problem can be cast and solved. Before entering in the technical details, an informal description of a typical situation in which the tuning problem arises is given in Section 3.1. In Section 3.2 we give the formal position of the problem and in Section 3.3 we discuss some possible extensions and variants. Section 3.4 summarizes and discusses the results presented in the chapter.