Length bias and search limitations in cartesian genetic programming

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

  author =       "Brian W. Goldman and William F. Punch",
  title =        "Length bias and search limitations in cartesian
                 genetic programming",
  booktitle =    "GECCO '13: Proceeding of the fifteenth annual
                 conference on Genetic and evolutionary computation
  year =         "2013",
  editor =       "Christian Blum and Enrique Alba and Anne Auger and 
                 Jaume Bacardit and Josh Bongard and Juergen Branke and 
                 Nicolas Bredeche and Dimo Brockhoff and 
                 Francisco Chicano and Alan Dorin and Rene Doursat and 
                 Aniko Ekart and Tobias Friedrich and Mario Giacobini and 
                 Mark Harman and Hitoshi Iba and Christian Igel and 
                 Thomas Jansen and Tim Kovacs and Taras Kowaliw and 
                 Manuel Lopez-Ibanez and Jose A. Lozano and Gabriel Luque and 
                 John McCall and Alberto Moraglio and 
                 Alison Motsinger-Reif and Frank Neumann and Gabriela Ochoa and 
                 Gustavo Olague and Yew-Soon Ong and 
                 Michael E. Palmer and Gisele Lobo Pappa and 
                 Konstantinos E. Parsopoulos and Thomas Schmickl and Stephen L. Smith and 
                 Christine Solnon and Thomas Stuetzle and El-Ghazali Talbi and 
                 Daniel Tauritz and Leonardo Vanneschi",
  isbn13 =       "978-1-4503-1963-8",
  pages =        "933--940",
  keywords =     "genetic algorithms, genetic programming, cartesian
                 genetic programming",
  month =        "6-10 " # jul,
  organisation = "SIGEVO",
  address =      "Amsterdam, The Netherlands",
  DOI =          "doi:10.1145/2463372.2463482",
  publisher =    "ACM",
  publisher_address = "New York, NY, USA",
  abstract =     "In this paper we examine how Cartesian Genetic
                 Programming's (CGP's) method for encoding directed
                 acyclic graphs (DAGs) and its mutation operator bias
                 the effective length of individuals as well as the
                 distribution of inactive nodes in the genome. We
                 investigate these biases experimentally using two CGP
                 variants as comparisons: Reorder, a method for
                 shuffling node ordering without effecting individual
                 evaluation, and DAG, a method for removing the concept
                 of node position. Experiments were performed on four
                 problems tailored to highlight potential search
                 limitations, with further testing on the 3-bit
                 multiplier problem.

                 Unlike previous work, our experiments show that CGP has
                 an innate parsimony pressure that makes it very
                 difficult to evolve individuals with a high percentage
                 of active nodes. This bias is particularly prevalent as
                 the length of an individual increases. Furthermore,
                 these problems are compounded by CGP's positional
                 biases which can make some problems effectively
                 unsolvable. Both Reorder and DAG appear to avoid these
                 problems and outperform Normal CGP on preliminary
                 benchmark testing. Finally, these new techniques
                 require more reasonable genome sizes than those
                 suggested in current CGP, with some evidence that
                 solutions are also more terse.",
  notes =        "Also known as \cite{2463482} GECCO-2013 A joint
                 meeting of the twenty second international conference
                 on genetic algorithms (ICGA-2013) and the eighteenth
                 annual genetic programming conference (GP-2013)",

Genetic Programming entries for Brian W Goldman William F Punch