Multi-Objective Methods for Tree Size Control

Created by W.Langdon from gp-bibliography.bib Revision:1.1934

@Article{dejong:2003:GPEM,
  author =       "Edwin D. {de Jong} and Jordan B. Pollack",
  title =        "Multi-Objective Methods for Tree Size Control",
  journal =      "Genetic Programming and Evolvable Machines",
  year =         "2003",
  volume =       "4",
  number =       "3",
  pages =        "211--233",
  month =        sep,
  keywords =     "genetic algorithms, genetic programming, variable size
                 representations, bloat, code growth, multi-objective
                 optimization, Pareto optimality, interpretability",
  ISSN =         "1389-2576",
  URL =          "http://www.cs.uu.nl/~dejong/publications/bloat.ps",
  URL =          "http://www.cs.uu.nl/~dejong/index.html#bloatgpem",
  doi =          "doi:10.1023/A:1025122906870",
  abstract =     "Variable length methods for evolutionary computation
                 can lead to a progressive and mainly unnecessary growth
                 of individuals, known as bloat. First, we propose to
                 measure performance in genetic programming as a
                 function of the number of nodes, rather than trees,
                 that have been evaluated. Evolutionary Multi-Objective
                 Optimisation (EMOO) constitutes a principled way to
                 optimise both size and fitness and may provide
                 parameterless size control. Reportedly, its use can
                 also lead to minimisation of size at the expense of
                 fitness. We replicate this problem, and an empirical
                 analysis suggests that multi-objective size control
                 particularly requires diversity maintenance.
                 Experiments support this explanation.

                 The multi-objective approach is compared to genetic
                 programming without size control on the 11-multiplexer,
                 6-parity, and a symbolic regression problem. On all
                 three test problems, the method greatly reduces bloat
                 and significantly improves fitness as a function of
                 computational expense. Using the FOCUS algorithm,
                 multi-objective size control is combined with active
                 pursuit of diversity, and hypothesised minimum-size
                 solutions to 3-, 4- and 5-parity are found. The
                 solutions thus found are furthermore easily
                 interpretable. When combined with diversity
                 maintenance, EMOO can provide an adequate and
                 parameterless approach to size control in variable
                 length evolution.",
  notes =        "Article ID: 5141122

                 Tue, 23 Mar 2004 01:22:36 +0100 See erratum in issue
                 5:1

                 Initial drop in size. 5-Parity given XOR!",
}

Genetic Programming entries for Edwin D de Jong Jordan B Pollack