The Effect of Extensive Use of the Mutation Operator on Generalization in Genetic Programming Using Sparse Data Sets

Created by W.Langdon from gp-bibliography.bib Revision:1.4496

  author =       "Wolfgang Banzhaf and Frank D. Francone and 
                 Peter Nordin",
  title =        "The Effect of Extensive Use of the Mutation Operator
                 on Generalization in Genetic Programming Using Sparse
                 Data Sets",
  booktitle =    "Parallel Problem Solving from Nature IV, Proceedings
                 of the International Conference on Evolutionary
  year =         "1996",
  editor =       "Hans-Michael Voigt and Werner Ebeling and 
                 Ingo Rechenberg and Hans-Paul Schwefel",
  series =       "LNCS",
  volume =       "1141",
  pages =        "300--309",
  address =      "Berlin, Germany",
  publisher_address = "Heidelberg, Germany",
  month =        "22-26 " # sep,
  publisher =    "Springer Verlag",
  keywords =     "genetic algorithms, genetic programming",
  ISBN =         "3-540-61723-X",
  DOI =          "doi:10.1007/3-540-61723-X_994",
  size =         "10 pages",
  abstract =     "Ordinarily, Genetic Programming uses little or no
                 mutation. Crossover is the predominant operator. This
                 study tests the effect of a very aggressive use of the
                 mutation operator on the generalisation performance of
                 our Compiling Genetic Programming System (CPGS). We ran
                 our tests on two benchmark classification problems on
                 very sparse training sets. In all, we performed 240
                 complete runs of population 3000 for each of the
                 problems, varying mutation rate between 5percent and
                 80percent. We found that increasing the mutation rate
                 can significantly improve the generalization
                 capabilities of GP. The mechanism by which mutation
                 affects the generalization capability of GP is not
                 entirely clear. What is clear is that changing the
                 balance between mutation and crossover effects the
                 course of GP training substantially - for example,
                 increasing mutation greatly extends the number of
                 generations for which the GP system can train before
                 the population converges.",
  notes =        " PPSN4
                 machine code GP CGPS used on IRIS, Gaussian 3D and
                 phoneme ELENA classification problems. Iris trivial. On
                 others best performance from 50/50 mix of crossover and

                 Answer extracted via designated hardware register. Stop
                 runs when destructive crossover falls below 10percent
                 (used as convergence indicator). Mutation giving rise
                 to more complex introns. GP premature convergence",
  affiliation =  "Dortmund University Department of Computer Science
                 Joseph-vonFraunhofer-Str. 20 44227 Dortmund Germany",

Genetic Programming entries for Wolfgang Banzhaf Frank D Francone Peter Nordin