Universal Consistency and Bloat in GP

Created by W.Langdon from gp-bibliography.bib Revision:1.1944

@Article{oai:hal.archives-ouvertes.fr:inria-00112840_v1,
  title =        "Universal Consistency and Bloat in {GP}",
  title_2 =      "Some theoretical considerations about Genetic
                 Programming from a Statistical Learning Theory
                 viewpoint",
  author =       "Sylvain Gelly and Olivier Teytaud and 
                 Nicolas Bredeche and Marc Schoenauer",
  journal =      "Revue d'Intelligence Artificielle",
  year =         "2006",
  volume =       "20",
  number =       "6",
  pages =        "805--827",
  note =         "Issue on New Methods in Machine Learning. Theory and
                 Applications",
  publisher =    "HAL - CCSd - CNRS",
  annote =       "Sylvain Gelly ",
  bibsource =    "OAI-PMH server at hal.archives-ouvertes.fr",
  contributor =  "Sylvain Gelly ",
  identifier =   "inria-00112840 (version 1)",
  oai =          "oai:hal.archives-ouvertes.fr:inria-00112840_v1",
  keywords =     "genetic algorithms, genetic programming, Computer
                 Science/Learning; Mathematics/Optimization and
                 Control",
  ISSN =         "0992-499X",
  URL =          "http://hal.inria.fr/docs/00/11/28/40/PDF/riabloat.pdf",
  URL =          "http://hal.inria.fr/inria-00112840/en/",
  URL =          "http://ria.revuesonline.com/article.jsp?articleId=8936",
  resume =       "Dans cet article, nous proposons une etude de la
                 Programmation Genetique (PG) du point de vue de la
                 theorie de l'Apprentissage Statistique dans le cadre de
                 la regression symbolique. En particulier, nous nous
                 sommes interesses a la consistence universelle en PG,
                 c'est-adire la convergence presque sure vers l'erreur
                 bayesienne a mesure que le nombre d'exemples augmente,
                 ainsi qu'au probleme bien connu en PG de la croissance
                 incontrolee de la taille du code (i.e. le {"}bloat{"}).
                 Les resultats que nous avons obtenus montrent d'une
                 part que l'on peut identifier plusieurs types de bloat
                 et d'autre part que la consistence universelle et
                 l'absence de bloat peuvent etre obtenues sous certaines
                 conditions. Nous proposons finalement une methode ad
                 hoc evitant justement le bloat tout en garantissant la
                 consistence universelle.",
  abstract =     "In this paper, we provide an analysis of Genetic
                 Programming (GP) from the Statistical Learning Theory
                 viewpoint in the scope of symbolic regression. Firstly,
                 we are interested in Universal Consistency, i.e. the
                 fact that the solution minimising the empirical error
                 does converge to the best possible error when the
                 number of examples goes to infinity, and secondly, we
                 focus our attention on the uncontrolled growth of
                 program length (i.e. bloat), which is a well-known
                 problem in GP. Results show that (1) several kinds of
                 code bloats may be identified and that (2) Universal
                 consistency can be obtained as well as avoiding bloat
                 under some conditions. We conclude by describing an ad
                 hoc method that makes it possible simultaneously to
                 avoid bloat and to ensure universal consistency.",
  notes =        "in english",
}

Genetic Programming entries for Sylvain Gelly Olivier Teytaud Nicolas Bredeche Marc Schoenauer