A benchmark generator of tree decomposition Mk landscapes

Abstract

We introduce the CliqueTreeMk algorithm to construct tree decomposition (TD) Mk Landscapes and to compute their global optimum efficiently. TD Mk Landscapes are well suited to serve as benchmark functions for blackbox genetic algorithms that are not given a priori the structural problem information as specified by the tree structure and their associated codomain fitness values. Specifically, for certain types of codomains the use of linkage learning might prove to be necessary in order to be able to solve these type of fitness functions.