- author = "Stephan M. Winkler",
- title = "Evolutionary System Identification: Modern Concepts and Practical Applications",
- publisher = "Trauner Verlag+Buchservice GmbH",
- year = "2009",
- number = "59",
- series = "Johannes Kepler University, Linz, Reihe C",
- keywords = "genetic algorithms, genetic programming",
- isbn13 = "978-3-85499-569-2",
-
URL = "
http://www.amazon.de/Evolutionary-System-Identification-Practical-Applications/dp/3854995695",
- broken = "http://www.trauner.at/buchdetail.aspx?artnr=20134591",
- abstract = "System identification denotes the data driven generation of mathematical models for systems; the result of a system identification algorithm consists in a mathematical description of the behaviour of the analysed system. Evolutionary computation is a subfield of computational intelligence that uses concepts inspired by natural evolution; one of the most famous evolutionary techniques is the genetic algorithm, a global optimisation technique using aspects inspired by evolutionary biology such as selection, recombination, mutation and inheritance. This thesis concentrates on evolutionary system identification techniques based on genetic programming (GP), an extension of the genetic algorithm: Mathematical expressions are produced by an evolutionary process that uses the given measurement data. The first part of this thesis describes theoretical concepts used in this work as well as our GP implementation for the HeuristicLab framework. Concepts for monitoring population dynamics during the execution of the GP process are also described; we here concentrate on genetic diversity and genetic propagation. The application of advanced selection principles and optimization stages is also explained as well as on-line and sliding window GP variants. The second part of this thesis summarises the results of system identification test series; the data sets used here include dynamic measurement data of mechatronical systems as well as classification benchmark problems. The results of these tests demonstrate the ability of this method to produce models of high quality for different kinds of machine learning problems, and also give insights into population dynamic processes that occur during the execution of a GP process.",
-
notes = "Dipl.-Ing. Dr. Stephan Winkler, Studium der Informatik
und Doktoratsstudium an der JKU in Linz. Bis 2006
wissenschaftlicher Mitarbeiter am LCM und am Institut
for Design und Regelung mechatronischer Systeme, danach
Anstellung im Rahmen des FWF Translational Research
Projekts L284 'GP-Based Techniques fort he Design
Virtual Sensors' an der FH Oeo, Campus Hagenberg. Ab
February 2009 Antritt einer Professur fuer
Bioinformatik an der FH Ooe.
Art. Nr. 20134591. See also \cite{Winkler:thesis}",
- size = "422 pages",
