The Effects of Randomly Sampled Training Data on Program Evolution

Created by W.Langdon from gp-bibliography.bib Revision:1.4216

  title =        "The Effects of Randomly Sampled Training Data on
                 Program Evolution",
  author =       "Brian J. Ross",
  institution =  "Dept. of Computer Science, Brock University",
  year =         "1999",
  type =         "Technical Report",
  number =       "CS-99-03",
  address =      "Canada",
  month =        nov,
  keywords =     "genetic algorithms, genetic programming",
  URL =          "",
  URL =          "",
  citeseer-references = "oai:CiteSeerPSU:178700; oai:CiteSeerPSU:127185;
                 oai:CiteSeerPSU:333851; oai:CiteSeerPSU:331862",
  annote =       "The Pennsylvania State University CiteSeer Archives",
  language =     "en",
  oai =          "oai:CiteSeerPSU:250158",
  rights =       "unrestricted",
  abstract =     "The effects of randomly sampled training data during
                 genetic programming is empirically investigated.
                 Sometimes the most natural, if not only, means of
                 characterizing the target behaviour for some problems
                 is to randomly sample training cases inherent to the
                 problems in question. A natural question to raise about
                 this strategy is, how deleterious is the randomly
                 sampling of training data to evolution performance?
                 Would such sampling reduce the evolutionary search to
                 hill climbing? We address these questions by
                 undertaking a suite of different GP experiments.
                 Various sampling strategies are used, such as different
                 training set sizes, single and multiple samples per
                 run, and manually derived {"}ideal distribution{"}
                 training sets. Both generational and steady--state
                 evolution are tested, in order to see if random
                 sampling particularly affects one or the other. Non--
                 evolutionary search strategies, such as hill climbing
                 and random search, are also used for comparison.
  notes =        "See also \cite{BRoss:2000:GECCO}",
  size =         "8 pages",

Genetic Programming entries for Brian J Ross