Ulrich Bodenhofer,
"Tuning Of Fuzzy Systems Using Genetic Algorithms"
, 3-1996, U. Bodenhofer. Tuning Of Fuzzy Systems Using Genetic Algorithms. Master's thesis, Johannes Kepler University Linz, March 1996.
Original Titel:
Tuning Of Fuzzy Systems Using Genetic Algorithms
Sprache des Titels:
Englisch
Englische Kurzfassung:
Since fuzzy logic has proven to be a very useful tool for
representing human knowledge by means of mathematical
expressions, the optimization of the involved parameters has been one of the most investigated problems in the theory of fuzzy expert systems. Typically, fuzzy systems have two components - a discrete one, the rules, and a continuous one on the other hand, the so-called fuzzy sets. Very many recent publications concern with the optimization of these two sets of parameters with genetic
algorithms (GAs). Genetic algorithms are optimization methods which are based on the mechanisms of natural evolution, such as selection, mutation, or sexual reproduction. The notion of genetic algorithms was
introduced approximately 25 years ago and turned out to be a very promising approach to the solution of many problems in artificial intelligence. During the last years the combination of fuzzy logic and GAs has come into fashion. Nevertheless, or better, for exactly that reason it is necessary to investigate this combination critically
and to expose the advantages and weaknesses objectively.
So far, we can distinguish between three classes of methods. The first one consists of approaches to the tuning of the first component, the fuzzy sets, which represent, in some sense, the semantic information of the rules. This mostly leads to a continuous optimization problems with real-valued parameters. The second class comprises methods for the discovery of optimal rulebases.
For these cases we typically get optimization problems in discrete, but not necessarily finite spaces. Last, we can collect all the methods, which do not fit in the first two classes,in a third group of methods. In particular, methods, where fuzzy sets and rules are tuned simultaneously belong to the third type. This thesis is intended to provide a profound introduction to both
fuzzy logic and genetic algorithms and to explore the possibilities to combine the two paradigms.
Erscheinungsmonat:
3
Erscheinungsjahr:
1996
Notiz zum Zitat:
U. Bodenhofer. Tuning Of Fuzzy Systems Using Genetic Algorithms. Master's thesis, Johannes Kepler University Linz, March 1996.
Forschungs-