"Analysing High Level Heuristics within a Graph Based Hyper-heuristic"
                Rong Qu 

This talk presents our work on analysing the high level heuristics within
a graph based hyper-heuristic (GHH) framework. Local search based methods
(Tabu Search and deepest descent method) are compared as the high level
heuristics within the GHH framework and the analysis of their neighborhood
structures and performance on the high level landscape is carried out. An
iterated deepest descent based GHH is developed based on the analysis
carried out. Experimental results on benchmark exam timetabling problems
demonstrate the simplicity and efficiency of the improved hyper-heuristic
approach. We suggest that the role the high level heuristics play is to
explore as larger as possible the landscape of the search space for both
the GHH and more general hyper-heuristics.