An Investigation and Extension of a Hyper-heuristic Framework

Three modifications to the framework within which hyper-heuristic approaches operate are presented. The first modification automates a self learning mechanism for updating the values of parameters in the choice function used by the controller. Second, a procedure for dynamically configuring a range of lowlevel heuristics is described. Third, in order to effectively use this range of low-level heuristics the controller is redesigned to form a hierarchy of sub-controllers. The second and third modifications improve the inflexibility associated with having a limited number of low-level heuristics available to the controller. Experiments are used to investigate features of the hyper-heuristic framework and the three modifications including comparisons with previously published results.