Recent Developments in Derivative-Free Multiobjective Optimisation

Abstract

In practical applications it is common to have several conflicting objective functions to optimize. Frequently, these functions are nondifferentiable or discontinuous, could be subject to numerical noise and/or be of black-box type, preventing the use of derivative-based techniques. In this paper we give an overview of some recent developments in Derivative-free Multiobjective Optimization. We introduce the basic concepts and ideas commonly considered for the algorithmic development in Multiobjective Optimization and review some recent classes of methods which do not make use of derivatives. In particular, we will focus on Direct Search Methods (DSM) of directional type and Evolutionary Multiobjective Optimization (EMO).