Hybridized Crossover-Based Search Techniques for Program Discovery

Created by W.Langdon from gp-bibliography.bib Revision:1.4420

  author =       "Una-May O'Reilly and Franz Oppacher",
  title =        "Hybridized Crossover-Based Search Techniques for
                 Program Discovery",
  institution =  "Santa Fe Institute",
  year =         "1995",
  number =       "95-02-007",
  address =      "1399 Hyde Park Road Santa Fe, New Mexico 87501-8943

  keywords =     "genetic algorithms, genetic programming",
  URL =          "http://www.santafe.edu/research/publications/workingpapers/95-02-007.ps",
  broken =       "http://www.ai.mit.edu/people/unamay/papers/xo-hybrid.ps",
  abstract =     "In this paper we address the problem of program
                 discovery as defined by Genetic Programming. We have
                 two major results: First, by combining a standard
                 crossover operator with two traditional single point
                 search algorithms (simulated annealing and stochastic
                 iterated hill climbing), we have solved some problems
                 with fewer fitness evaluations and a greater
                 probability of a success than Genetic Programming.
                 Second, we have managed to enhance Genetic Programming
                 by hybridizing it with the simple scheme of hill
                 climbing from a few individuals, at a fixed interval of
                 generations. The new hillclimbing component has two
                 options for generating candidate solutions: mutation or
                 crossover. When it uses crossover, mates are either
                 randomly selected or are individually drawn from the
                 population at large, or are drawn from a pool of
                 fittest individuals. The population pool option has
                 proved superior thus indicating that a combination of
                 population-based evolution and greedy exploitation of a
                 single individual has merit.

  notes =        "If you want the paper version contact SFI
                 (mat@santafe.edu) or contact una-may for a postscript
                 uuencoded version. All comments are welcome. Contact me
                 with unamay@santafe.edu.

                 The unabridged version of this paper is Santa Fe
                 Institute Working Paper: 95-02-007. An abridged (6
                 page) version is to appear in the proceedings of the
                 1995 World Conference on Evolutionary Computation held
                 in Perth, Australia, December 1-3, 1995.",
  size =         "11 pages",

Genetic Programming entries for Una-May O'Reilly Franz Oppacher