Neural Network Synthesis using Cellular Encoding and the Genetic Algorithm.   [NN] [NS] [CE] [GA]

by

Gruau, F.

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: 1994
Keywords:genetic algorithms, genetic programming
Abstract:
Artificial neural networks [ANN] [NN] used to be considered only as a machine that learns using small modifications of internal parameters. Now this is changing. Such learning method do not allow to generate big neural networks [NN] for solving real world problems. [RWP] This thesis defends the following three points: (1) The key word to go out of that dead-end is "modularity". (2) The tool that can generate modular neural networks [NN] is cellular encoding. [CE] (3) The optimization algorithm adapted to the search of cellular codes is the genetic algorithm. [GA] The first point is now a common idea. A modular neural network [NN] means a neural network [NN] that is made of several sub-networks, arranged in a hierarchical way. For example, the same sub-network can be repeated. This thesis encompasses two parts. The first part demonstrates the second point. Cellular encoding [CE] is presented as a machine language for neural networks, [NN] with a theoretical basis (it is a parallel graph grammar that checks a number of properties) and a compiler of high level language. The second part of the thesis shows the third point. Application of genetic algorithm to [GA] the synthesis of neural networks [NN] using cellular encoding [CE] is a new technology. This technology can solve problems that were still unsolved with neural networks. [NN] It can automatically and dynamically decompose a problem into a hierarchy of sub-problems, and generate a neural network [NN] solution to the problem. The structure of this network is a hierarchy of sub-networks that reflects the structure of the problem. The technology allows to experience new scientific domains like the interaction between learning and evolution, or the set up of learning algorithms that suit the GA.
Notes:
URL(s):(G)zipped postscript

Review item:

Mark as doublet (will be reviewed)

Print entry



BibTex:
@PhdThesis{Gruau:1994:thesis,
  author =       "F. Gruau",
  title =        "Neural Network Synthesis using Cellular Encoding and
                 the Genetic Algorithm.",
  school =       "Laboratoire de l'Informatique du Parallilisme, Ecole
                 Normale Supirieure de Lyon",
  year =         "1994",
  address =      "France",
  keywords =     "genetic algorithms, genetic programming",
  URL =          "ftp://lip.ens-lyon.fr/pub/Rapports/PhD/PhD94-01-E.ps.Z",
  url2 =         "ftp://lip.ens-lyon.fr/pub/Rapports/PhD/PhD94-01-F.ps.Z",
  size =         "151 pages",
  abstract =     "Artificial neural networks used to be considered only
                 as a machine that learns using small modifications of
                 internal parameters. Now this is changing. Such
                 learning method do not allow to generate big neural
                 networks for solving real world problems. This thesis
                 defends the following three points:

                 (1) The key word to go out of that dead-end is
                 {"}modularity{"}. (2) The tool that can generate
                 modular neural networks is cellular encoding. (3) The
                 optimization algorithm adapted to the search of
                 cellular codes is the genetic algorithm.

                 The first point is now a common idea. A modular neural
                 network means a neural network that is made of several
                 sub-networks, arranged in a hierarchical way. For
                 example, the same sub-network can be repeated. This
                 thesis encompasses two parts.

                 The first part demonstrates the second point. Cellular
                 encoding is presented as a machine language for neural
                 networks, with a theoretical basis (it is a parallel
                 graph grammar that checks a number of properties) and a
                 compiler of high level language. The second part of the
                 thesis shows the third point. Application of genetic
                 algorithm to the synthesis of neural networks using
                 cellular encoding is a new technology. This technology
                 can solve problems that were still unsolved with neural
                 networks. It can automatically and dynamically
                 decompose a problem into a hierarchy of sub-problems,
                 and generate a neural network solution to the problem.
                 The structure of this network is a hierarchy of
                 sub-networks that reflects the structure of the
                 problem. The technology allows to experience new
                 scientific domains like the interaction between
                 learning and evolution, or the set up of learning
                 algorithms that suit the GA.",
  notes =        "

                 ",
}