Solving exam timetabling using distributed evolutionary computation

In this paper we describe specific exam timetabling problem encountered at our institution. We enumerate and briefly explain population based evolutionary computation algorithms that we implemented and then focus on coarse-grained algorithm parallelization. Parallelization is accomplished by employing a computer network in which separate populations exchange best solutions. Solution exchange is guided by various migration parameters such as network exchange topology. Proposed topologies are tested and experimental results are discussed. We conclude that usage of a computer network, exchange topologies, and large amount of processor time enabled us to find a good-quality exam timetable.