Introns in Nature and in Simulated Structure Evolution

Created by W.Langdon from gp-bibliography.bib Revision:1.2031

@InProceedings{nordin:1997:insse,
  author =       "P. Nordin and W. Banzhaf and F. D. Francone",
  title =        "Introns in Nature and in Simulated Structure
                 Evolution",
  booktitle =    "Bio-Computation and Emergent Computation",
  year =         "1997",
  editor =       "Dan Lundh and Bjorn Olsson and Ajit Narayanan",
  address =      "Skovde, Sweden",
  month =        "1-2 " # sep,
  publisher =    "World Scientific Publishing",
  keywords =     "genetic algorithms, genetic programming",
  ISBN =         "981-02-3262-4",
  URL =          "http://coblitz.codeen.org:3125/citeseer.ist.psu.edu/cache/papers/cs/598/http:zSzzSzls11-www.informatik.uni-dortmund.dezSzpeoplezSzbanzhafzSzskovde_final.pdf/nordin97introns.pdf",
  URL =          "http://citeseer.ist.psu.edu/nordin97introns.html",
  URL =          "http://www.wspc.com.sg/books/lifesci/3593.html",
  abstract =     "In this study we measure the compression of
                 information in a simulated evolutionary system. We do
                 the investigation taking introns in the genome into
                 account. We mainly investigate evolution of linear
                 computer code but also present results from evolution
                 of tree structures as well as messy genetic algorithms.
                 The size of solutions is an important property of any
                 system trying to learn or adapt to its environment. The
                 results show significant compression or constant size
                 of exons during evolution - in contrast to the rapid
                 growth of overall size. Our conclusion is that an
                 built-in pressure towards low-complexity solutions is
                 measurable in several simulated evolutionary systems
                 which may account for the robust adaptation shown by
                 these systems.",
  notes =        "BCEC97

                 We change one change each in struction to a NoOperation
                 (NOP) and if this doesn't make any difference for any
                 of the fitness cases then we call it an (first order)
                 intron. For the tree-based-GP we look at what values
                 that come in to each node and what values that
                 propagates upwards from each node. If one value into
                 the node is the same as one value out from the node
                 *for all fitness cases* then the other sub-trees under
                 the node are introns.

                 The tree run was symbolic regression of a 3rd order
                 polynomial",
}

Genetic Programming entries for Peter Nordin Wolfgang Banzhaf Frank D Francone