Distributed choice function hyper-heuristics for timetabling and scheduling

Abstract

This paper investigates an emerging class of search algorithms, in which high-level domain independent heuristics, called hyper-heuristics, iteratively select and execute a set of application specific but simple search moves, called low-level heuristics, working toward achieving improved or even optimal solutions. Parallel architectures have been designed and evaluated. Results based on a university timetabling problem show an important relationship between performance, algorithm software and hardware implementation. © Springer-Verlag Berlin Heidelberg 2005.