Classification as Clustering: A Pareto Cooperative-Competitive GP Approach

Created by W.Langdon from gp-bibliography.bib Revision:1.4524

  author =       "Andrew R. McIntyre and Malcolm I. Heywood",
  title =        "Classification as Clustering: A Pareto
                 Cooperative-Competitive GP Approach",
  journal =      "Evolutionary Computation",
  year =         "2011",
  volume =       "19",
  number =       "1",
  pages =        "137--166",
  month =        "Spring",
  keywords =     "genetic algorithms, genetic programming, MOGA,
  ISSN =         "1063-6560",
  DOI =          "doi:10.1162/EVCO_a_00016",
  size =         "30 pages",
  abstract =     "Intuitively population based algorithms such as
                 Genetic Programming provide a natural environment for
                 supporting solutions that learn to decompose the
                 overall task between multiple individuals, or a team.
                 This work presents a framework for evolving teams
                 without recourse to pre-specifying the number of
                 cooperating individuals. To do so, each individual
                 evolves a mapping to a distribution of outcomes that,
                 following clustering, establishes the parametrisation
                 of a (Gaussian) local membership function. This gives
                 individuals the opportunity to represent subsets of
                 tasks, where the overall task is that of classification
                 under the supervised learning domain. Thus, rather than
                 each team member represent an entire class, individuals
                 are free to identify unique subsets of the overall
                 classification task. The framework is supported by
                 techniques from Evolutionary Multi-objective
                 Optimisation (EMO) and Pareto competitive coevolution.
                 EMO establishes the basis for encouraging individuals
                 to provide accurate yet non-overlaping behaviours;
                 whereas competitive coevolution provides the mechanism
                 for scaling to potentially large unbalanced data sets.
                 Benchmarking is performed against recent examples of
                 non-linear SVM classifiers over twelve UCI data sets
                 with between 150 and 200,000 training instances.
                 Solutions from the proposed Coevolutionary
                 Multi-objective GP framework appear to provide a good
                 balance between classification performance and model
                 complexity, especially as the data set instance count

Genetic Programming entries for Andrew R McIntyre Malcolm Heywood