Automated design of metaheuristic algorithms

Abstract

The design and development of metaheuristic algorithms can be time-consuming and difficult for a number of reasons including the complexity of the problems being tackled, the large number of degrees of freedom when designing an algorithm and setting its numerical parameters, and the difficulties of algorithm analysis due to heuristic biases and stochasticity. Traditionally, this design and development has been done through a manual, labor-intensive approach guided mainly by the expertise and intuition of the algorithm designer. In recent years, a number of automatic algorithm configuration methods have been developed that are able to effectively search large and diverse parameter spaces. They have been shown to be very successful in identifying high-performing algorithm designs and parameter settings. In this chapter, we review the recent advances in addressing automatic metaheuristic algorithm design and configuration. We describe the main existing automatic algorithm configuration techniques and discuss some of the main uses of such techniques, ranging from the mere optimization of the performance of already developed metaheuristic algorithms to their pivotal role in modifying the way metaheuristic algorithms will be designed and developed in the future.