2001

 nach oben zur Jahresübersicht

• Nicole Drechsler, Frank Schmiedle, Daniel Große, Rolf Drechsler
  Heuristic Learning based on Genetic Programming
  2001 EuroGP, Springer Verlag, Band: 2038, Seiten: 1 - 10
• Frank Schmiedle, Nicole Drechsler, Daniel Große, Rolf Drechsler
  Priorities in Multi-Objective Optimization for Genetic Programming
  2001 Conf. on Genetic and Evolutionary Computation, Seiten: 129 - 136
  » Kurzfassung anzeigen« Kurzfassung verbergen
  Kurzfassung

  A new technique for multi-objective optimization is presented that allows to include priorities. But in contrast to previous techniques they can be included very easily and do not require much user interaction. The new approach is studied from a theoretical and practical point of view. The main differences to existing methods, like relation dominate and favor, are discussed. An experimental study of applying priorities in heuristics learning based on Genetic Programming (GP) is described. The experiments confirm the advantages presented in comparison to several other techniques.
• Frank Schmiedle, Daniel Große, Rolf Drechsler, Bernd Becker
  Too Much Knowledge Hurts: Acceleration of Genetic Programs for Learning Heuristics
  2001 Int'l Conf. on Computational Intelligence, Band: 2206, Seiten: 479 - 491
  » Kurzfassung anzeigen« Kurzfassung verbergen
  Kurzfassung

  Among many other applications, evolutionary methods have been used to develop heuristics for several optimization problems in VLSI CAD in recent years. Although learning is performed according to a set of training benchmarks, it is most important to generate heuristics that have a good generalization behaviour and hence are well suited to be applied to unknown examples. Besides large runtimes for learning, the major drawback of these approaches is that they are very sensitive to a variety of parameters for the learning process. In this paper, we study the impact of different parameters, like e.g. stopping conditions, on the quality of the results for learning heuristics for BDD minimization. If learning takes too long, the developed heuristics become too specific for the set of training examples and in that case results of application to unknown problem instances deteriorate. It will be demonstrated here that runtime can be saved while even improving the generalization behaviour of the heuristics.