Scalability of Genetic Programming and Probabilistic Incremental Program Evolution

Created by W.Langdon from gp-bibliography.bib Revision:1.4340

  title =        "Scalability of Genetic Programming and Probabilistic
                 Incremental Program Evolution",
  note =         "Comment: Submitted to GECCO-2005",
  author =       "Radovan Ondas and Martin Pelikan and Kumara Sastry",
  year =         "2005",
  month =        feb # "~07",
  bibsource =    "OAI-PMH server at",
  oai =          "",
  URL =          "",
  URL =          "",
  howpublished = "arXiv",
  keywords =     "genetic algorithms, genetic programming, PIPE, Neural
                 and Evolutionary Computing, Artificial Intelligence",
  size =         "8 pages",
  abstract =     "This paper discusses scalability of standard genetic
                 programming (GP) and the probabilistic incremental
                 program evolution (PIPE). To investigate the need for
                 both effective mixing and linkage learning, two test
                 problems are considered: ORDER problem, which is rather
                 easy for any recombination-based GP, and TRAP or the
                 deceptive trap problem, which requires the algorithm to
                 learn interactions among subsets of terminals. The
                 scalability results show that both GP and PIPE scale up
                 polynomially with problem size on the simple ORDER
                 problem, but they both scale up exponentially on the
                 deceptive problem. This indicates that while standard
                 recombination is sufficient when no interactions need
                 to be considered, for some problems linkage learning is
                 necessary. These results are in agreement with the
                 lessons learned in the domain of binary-string genetic
                 algorithms (GAs). Furthermore, the paper investigates
                 the effects of introducing utnnecessary and irrelevant
                 primitives on the performance of GP and PIPE.",
  notes =        "see GECCO poster \cite{1068310}


                 Page 3 {"}Interactions between each node and its
                 context are ignored. That is why it [PIPE] can be
                 expected ... inferior results. ... if ... components
                 are ... independent, PIPE should work great.{"}

                 ORDER onemax like. TRAP deceptive, GP-difficult. (k=3
                 d=1). JOIN NEG_JOIN functions, JUNK terminals

                 page 5 {"}Population size determined by bisection

Genetic Programming entries for Radovan Ondas Martin Pelikan Kumara Sastry