Evolver: Meta-optimizing multi-objective metaheuristics.

Abstract

Evolver is a tool based on the formulation of the automatic configuration and design of multi-objective metaheuristics as a multi-objective optimization problem that can be solved by using the same kind of algorithms; i.e., we are applying a meta-optimization approach. Evolver provides highly configurable implementations of representative multi-objective solvers which can be automatically configured from a number of multi-objective problems used as the training set and a list of quality indicators which are the objectives to be optimized. Our tool is based on the jMetal framework, so a large number of existing algorithms can be used as meta-optimizers. A graphical user interface allows scientists to easily define auto-configuration scenarios, thus simplifying the complex process of finding high-quality algorithm settings.