Evolving better representations through selective genome growth

by

Altenberg, L.

Literature search on Evolutionary ComputationBBase ©1999-2013, Rasmus K. Ursem
     Home · Search · Adv. search · Authors · Login · Add entries   Webmaster
Note to authors: Please submit your bibliography and contact information - online papers are more frequently cited.

Info: Proceedings of the 1st IEEE Conference on Evolutionary Computation (Conference proceedings), 1994, p. 182-187
Keywords:genetic algorithms, genetic programming
Abstract:
The choice of how to represent the search space [SS] for a genetic algorithm (GA) [GA] is critical to the GA's performance. Representations are usually engineered by hand and fixed for the duration of the GA run. Here a new method is described in which the degrees of freedom of the representation --- i.e. the genes -- are increased incrementally. The phenotypic effects of the new genes are randomly drawn from a space of different functional effects. Only those genes that initially increase fitness are kept. The genotype-phenotype map [GM] that results from this selection during the constructional of the genome allows better adaptation. This effect is illustrated with the NK landscape model. The resulting genotype-phenotype maps are much less epistatic than generic maps would be. They have extremely low values of ``K'' --- the number [TNO] of fitness components affected by each gene. Moreover, these maps are exquisitely tuned to the specifics of the random fitness functions, [FF] and achieve fitnesses many standard deviations above generic NK landscapes [NL] with the same \gp\ maps. The evolved maps create adaptive landscapes that are much smoother than generic NK landscapes [NL] ever are. Thus a caveat should be made when making arguments about the applicability of generic properties of complex systems to evolved systems. This method may help to solve the problem of choice of representations in genetic algorithms. [GA] Copyright 1996 Lee Altenberg
Notes:
URL(s):(G)zipped postscript
HTML

Review item:

Mark as doublet (will be reviewed)

Print entry




BibTex:
@InProceedings{Altenberg:1994EBR,
  author =       "Lee Altenberg",
  year =         "1994",
  pages =        "182--187",
  title =        "Evolving better representations through selective
                 genome growth",
  booktitle =    "Proceedings of the 1st IEEE Conference on Evolutionary
                 Computation",
  publisher =    "IEEE",
  address =      "Orlando, Florida, USA",
  month =        "27-29 " # jun,
  publisher_address = "Piscataway, NJ, USA",
  volume =       "1",
  keywords =     "genetic algorithms, genetic programming",
  URL =          "ftp://ftp.mhpcc.edu/pub/incoming/altenberg/LeeEBR.ps.Z",
  url_2 =        "http://pueo.mhpcc.edu/~altenber/PAPERS/LeeEBR.html",
  abstract =     "The choice of how to represent the search space for a
                 genetic algorithm (GA) is critical to the GA's
                 performance. Representations are usually engineered by
                 hand and fixed for the duration of the GA run. Here a
                 new method is described in which the degrees of freedom
                 of the representation --- i.e. the genes -- are
                 increased incrementally. The phenotypic effects of the
                 new genes are randomly drawn from a space of different
                 functional effects. Only those genes that initially
                 increase fitness are kept. The genotype-phenotype map
                 that results from this selection during the
                 constructional of the genome allows better adaptation.
                 This effect is illustrated with the NK landscape model.
                 The resulting genotype-phenotype maps are much less
                 epistatic than generic maps would be. They have
                 extremely low values of ``K'' --- the number of fitness
                 components affected by each gene. Moreover, these maps
                 are exquisitely tuned to the specifics of the random
                 fitness functions, and achieve fitnesses many standard
                 deviations above generic NK landscapes with the same
                 \gp\ maps. The evolved maps create adaptive landscapes
                 that are much smoother than generic NK landscapes ever
                 are. Thus a caveat should be made when making arguments
                 about the applicability of generic properties of
                 complex systems to evolved systems. This method may
                 help to solve the problem of choice of representations
                 in genetic algorithms.

                 Copyright 1996 Lee Altenberg",
  notes =        "

                 ",
}