An Investigation of Hyper-Heuristic Methods: A New Generation of Problems-Solvers
01/11/2000 - 31/07/2004
research seeks to investigate and develop heuristics that choose
heuristics (hyper-heuristics). Problems which may appear similar can
require different heuristics to produce good quality solutions. For
example, in stock cutting where space is at a premium, large shapes can
be effectively placed by considering those first. In another stock
cutting problem where space is not such an issue, this strategy produces
poorer solutions. For this problem domain a two phase approach is
likely to produce good results. In the first phase one heuristic
prioritises the order in which shapes are considered and phase two is
used to place each shape. In order to explore the general idea further,
we need a wide-ranging investigation that develops heuristics that
selects ,other heuristics and applies this hyper-heuristic idea to many
different problem domains. A successful outcome of this ambitious but
achievable project would provide a significant step forward in the way
these difficult problems are handled and also increase our understanding
as to the nature of these problems and how best to solve them.
The project made significant advances in discovering new automated methods for combining heuristics to solve bin-packing problems, outperforming results obtained by any individual heuristic.
Advances were made using learning classifier systems and evolutionary algorithms to learn characteristics of problems and partial solutions, and map those characteristics to suitable heuristics.
The project results in many publications in conferences and as book-chapters,, including a best paper award at GECCO 02 for "Hyper-heurstics:
learning to combine simple heuristics in bin-packing problems."
An Investigation of Hyper-Heuristic Methods: A New Generation of Problems-Solvers is a Research Councils project funded by EPSRC.
Carried out in collaboration with and others.
For further information please refer to .
The Bio-Inspired Algorithms group within the Centre for Algorithms, Visualisation and Evolving Systems is a large and thriving group with interests in nature-inspired computing that include Evolutionary Computing, Hyper-Heuristics, Artificial Immune Systems and Swarm Intelligence.
Director of CEC
+44 131 455 2783
+44 131 455