Track: Artificial Intelligence

Abstract

Metaheuristics have many parameters and a good selection of that parameter values can increase efficiency and effectiveness of these algorithms when solving optimization problems. A considerable number of automated parameter tuning methods have been developed in the last few years. Recently developed Instance-specific Parameter Tuning Strategies (IPTS) considers the interaction between measurable test problem characteristics and algorithm-specific parameter values when developing a tuning strategy.  Although algorithm configuration has a very wide literature, there are rather few studies suggested on instance-specific parameter tuning and there is not a comprehensive literature survey on this field. In this study we will first give a definition of the algorithm configuration problem and will give an overview about open access studies in the field of Instance-specific Parameter Tuning Strategies (IPTS). We will survey the proposed methods on this topic and discuss the advantages and disadvantages of these methods as a conclusion in this study. This is first study that surveys the proposed methods in the field of Instance-specific Parameter Tuning Strategies (IPTS).