Experimental Supplements to the Theoretical Analysis of EAs on Problems from Combinatorial Optimization

It is typical for the EA community that theory follows experiments. Most theoretical approaches use some model of the considered evolutionary algorithm (EA) but there is also some progress where the expected optimization time of EAs is analyzed rigorously. There are only three well-known problems of combinatorial optimization where such an approach has been performed for general input instances, namely minimum spanning trees, shortest paths, and maximum matchings. The theoretical results are asymptotic ones and several questions for realistic dimensions of the search space are open. We supplement the theoretical results by experimental ones. Many hypotheses are confirmed by rigorous statistical tests.

• Type: Conference
• Year: 2004
• Where: PPSN
• Authors: Patrick Briest, Dimo Brockhoff, Bastian Degener, Matthias Englert, Christian Gunia, Oliver Heering, Thomas Jansen, Michael Leifhelm, Kai Plociennik, Heiko Röglin, Andrea Schweer, Dirk Sudholt, Stefan Tannenbaum, Ingo Wegener
• Keywords: Distributed And Parallel Computing, General Input Instances, Minimum Spanning Trees, Most Theoretical Approaches, PPSN 2004