Batched Mode Hyper-heuristics

A common form of a hyper-heuristic is a method that controls a search process which uses neighbourhood operators. There have many studies showing that hyper-heuristics are reusable for solving unseen problem instances not only from a particular domain but also different problem domains without requiring any change. However, generally hyper-heuristics have been considered as part of time-contract algorithms, i.e. they are given a fixed execution time, and also used to solve each each instance separately. This paper considers the potential gains and challenges of a hyper-heuristic being able to treat a set of instances as a batch, to be completed within an overall joint execution time, but with the hyper-heuristic free to make its own decision as to how to divide the computational effort between the individual instances. Using the standard CHeSC benchmarks, we show the wide variation in runtimes that occur, and give evidence that this results in a batched mode having the potential for significant gains.