Stream on Hyper-heuristics

the 22nd Conference on International Federation of Operational Research Societies (IFORS 2020), June 21-26 2020, Seoul/Korea

Call for Abstracts (Max. 1500 characters)

• Session 1: Selection Hyper-heuristics (Submission Code: 3298ef0c)

• Session 2: Generation Hyper-heuristics (Submission Code: e964da30)

Hyper-heuristics are problem-independent generic solvers which have been successfully applied to a wide range of combinatorial search problems both from academia and real-world, such as timetabling, scheduling, routing, rostering, cutting and packing. The studies on this field is mainly considered under two categories (Figure 1), namely Selection and Generation Hyper-heuristics. Selection Hyper-heuristics operate by automatically choosing (low-level) heuristics from an existing heuristic set while the latter type focuses on generating heuristics from scratch based on predefined components. These hyper-heuristics can have certain learning capabilities by incorporating Offline and Online learning. Offline refers to learning before a hyper-heuristic run, mostly in the form of un-/supervised learning. Online is about learning while a problem (~instance) is being solved, likely to be based on reinforcement learning. It is also possible to design hyper-heuristics without learning. Besides the learning aspect, the type of heuristics may differ as constructive and perturbative heuristics.

The aim of this stream is to gather researchers studying hyper-heuristics to share their research on all the aforementioned hyper-heuristic variations as well as the strategies developed to support hyper-heuristics.

Organizers