Wiesmann, Dirk2004-12-072004-12-0720002001-10-16http://hdl.handle.net/2003/538610.17877/DE290R-5638Evolutionary programming has originally been proposed for the breeding of finite state automata. The mutation operator is working directly on the graph structure of the automata. In this paper we introduce variation operators based on the automatons input/output behavior rather than its structure. The operators are designed to make use of additional information based on a ranking of states as well as a problem-specific metric which enhances the search process.enUniversität DortmundReihe Computational Intelligence ; 82004report