INCOP: An open library for incomplete combinatorial optimization

Abstract

We present a new library, INCOP, which provides incomplete algorithms for optimizing combinatorial problems. This library offers lo cal search methods such as simulated annealing, tabu search as well as a population based method, Go With the Winners. Several problems have been encoded, including Constraint Satisfaction Problems, graph coloring, frequency assignment. INCOP is an open C++ library. The user can easily add new algorithms and encode new problems. The neighborhood management has been care fully studied. First, an original parameterized move selection allows us to easily implement most of the existing meta-heuristics. Second, differ ent levels of incrementality can be specified for the configuration cost computation, which highly improves efficiency. INCOP has shown great performances on well-known benchmarks. The challenging flat300_28 graph coloring instance has been colored in 30 colors for the first time by a standard Metropolis algorithm. © Springer-Verlag 2003.