PM-DGP A Distributed Genetic Programming Framework

Created by W.Langdon from gp-bibliography.bib Revision:1.4221

  author =       "P. G. M. {van der Meulen}",
  title =        "{PM-DGP} A Distributed Genetic Programming Framework",
  school =       "Laboratory for Signals \& Systems, Department of
                 Electrical Engineering",
  year =         "2001",
  type =         "Master's thesis",
  address =      "University of Twente, The Netherlands",
  month =        oct,
  keywords =     "genetic algorithms, genetic programming",
  URL =          "",
  size =         "89 pages",
  abstract =     "In this report the design, implementation and usage of
                 PM-DGP is described. PM-DGP is a fully object-oriented
                 framework, written in C++, to aid in the implementation
                 of genetic programming (GP) problems. It is the result
                 of the search for a GP environment that allows a
                 programmer to concentrate on writing the fitness
                 function and the problem specific nodes he needs and
                 let the system take care of the rest, including the
                 distribution of the fitness evaluation. The system is
                 freely available as it is released under the GNU public

                 The framework supports important additions to GP like
                 automatically defined functions and random constants.
                 Once a problem has been implemented using PM-DGP the
                 time consuming task of fitness evaluation can be
                 distributed using the idle time of networked computers
                 running Microsoft Windows or Linux using a GUI server
                 and clients. The system is designed to be flexible and
                 extensible. Many aspects of a GP run, the nodes, node
                 sets, result type and genetic algorithm, are
                 configurable at run time using a simple textual
                 configuration file.

                 The system offers a flexible yet efficient
                 object-oriented genome interpreter that can work with
                 different result types and uses a prefix coding to
                 store its programs. When distributing the fitness
                 evaluation the genome interpreter has to be sent to a
                 client over the network only once after which the
                 programs can be sent using only one to two bytes per
                 node. Several example implementations of GP problems
                 are included: symbolic regression, parity, artificial
                 ant, royal tree and edge detector. An extensive
                 tutorial showing how to implement the lawn mower
                 problem is included in this report.",
  notes =        "Graduation Committee: Ir. A.M. Bazen (EL-S&S)
                 (supervisor) Dr. ir. S.H. Gerez (EL-S&S) (supervisor)
                 H.J. Kip, Btw (Nedap) Ir. A. Kuip (Nedap) Dr. M.Poel
                 (INF-TT) Prof. dr. ir. C.H. Slump (EL-S&S) Date:
                 11-Oct-2001 Report number: EL-S&S


Genetic Programming entries for P G M van der Meulen