Worst-Case Optimal Algorithms

Abstract

The question of the possibility to construct an optimal (in some sense) algorithm is of importance to the theory of multi-objective optimization similarly to any other theory of algorithmically solvable problems. In this chapter, we aim at finding a worst-case optimal approximation of the Pareto optimal set for multi-objective optimization problems, where the convexity of objective functions is not assumed. The class of Lipschitz functions is chosen as a model of objective functions since that model is one of the simplest and best researched models of global optimization [87]. Worst-case optimal algorithms are constructed for the cases of passive (non-adaptive) and sequential (adaptive) search in [249]. These results are the generalization to the multi-objective case of the results by Sukharev who investigated the worst-case optimal single-objective optimization algorithms in [210, 211].