Accelerating Evolutionary Algorithms Using Fitness Function Models

D. Büche, N. N. Schraudolph, and P. Koumoutsakos. Accelerating Evolutionary Algorithms Using Fitness Function Models. In Genetic and Evolutionary Computation Conference Workshop Program, pp. 166–169, AAAI, Chicago, 2003.
Latest version

Download

188.8kB	89.5kB	96.9kB

Abstract

An optimization procedure using empirical models as an approximation of expensive functions is presented. The model is trained on the current set of evaluated solutions and can be used to search for promising solutions. These solutions are then evaluated on the expensive function. The resulting iterative procedure is analyzed and compared on test problems and shows fast and robust convergence. In particular, implementation issues like local modeling of the data, parallelization of the function evaluation and avoiding premature convergence are discussed.

BibTeX Entry

@inproceedings{BueSchKou03,
     author = {Dirk B\"uche and Nicol N. Schraudolph
               and Petros Koumoutsakos},
      title = {\href{http://nic.schraudolph.org/pubs/BueSchKou03.pdf}{
               Accelerating Evolutionary Algorithms Using
               Fitness Function Models}},
      pages = {166--169},
     editor = {Alwyn M. Barry},
  booktitle = {Genetic and Evolutionary Computation Conference
               Workshop Program},
  publisher = {AAAI},
    address = {Chicago},
       year =  2003,
   b2h_type = {Other},
  b2h_topic = {Evolutionary Algorithms},
   b2h_note = {<a href="b2hd-BueSchKou05.html">Latest version</a>},
   abstract = {
    An optimization procedure using empirical models as an approximation
    of expensive functions is presented.  The model is trained on the
    current set of evaluated solutions and can be used to search for
    promising solutions.  These solutions are then evaluated on the
    expensive function.  The resulting iterative procedure is analyzed
    and compared on test problems and shows fast and robust convergence.
    In particular, implementation issues like local modeling of the data,
    parallelization of the function evaluation and avoiding premature
    convergence are discussed.
}}