Memes, self-generation and nurse rostering

This paper presents an empirical study on memetic algorithms in two parts. In the first part, the details of the memetic algorithm experiments with a set of well known benchmark functions are described. In the second part, a heuristic template is introduced for solving timetabling problems. Two adaptive heuristics that utilize a set of constraint-based hill climbers in a co-operative manner are designed based on this template. A hyper-heuristic is a mechanism used for managing a set of low-level heuristics. At each step, an appropriate heuristic is chosen and applied to a candidate solution. Both adaptive heuristics can be considered as hyper-heuristics. Memetic algorithms employing each hyper-heuristic separately as a single hill climber are experimented on a set of randomly generated nurse rostering problem instances. Moreover, the standard genetic algorithm and two self-generating multimeme memetic algorithms are compared to the proposed memetic algorithms and a previous study.