The Physics behind genetic programming

Created by W.Langdon from gp-bibliography.bib Revision:1.3973

@TechReport{werner:2002:gpphys,
  author =       "James Cunha Werner",
  title =        "The Physics behind genetic programming",
  institution =  "London South Bank University",
  year =         "2001",
  address =      "SCISM, South Bank University, 103 Borough Road, London
                 SE1 0AA",
  keywords =     "genetic algorithms, genetic programming, Calculus of
                 Variations, Euler-Lagrange Equation",
  URL =          "http://www.geocities.com/jamwer2002/gpphys.pdf",
  size =         "8 pages",
  abstract =     "the historic scenery of calculus of variations (CV),
                 one of the central tools of theoretical physics, and
                 its relationship with genetic programming (GP)
                 algorithms, a search method with would be considered a
                 numerical solution for the method of
                 variations.

                 Conclusion. This paper establishes a relationship
                 between the CV and GP as its numerical methods. The
                 central goal of CV is determining the functional that
                 attend to some constraints solving fundamentals
                 differential equations by analytical methods while GP
                 try to obtain the solution applying genetic operators
                 in tree coded chromosomes. The differential
                 displacement in analytical solution assumes the format
                 of a change into the functional structure through the
                 application of crossover and mutation operators. The
                 action integral has its similar in the fitness
                 function, with in the same way is obtained during all
                 solution interval time. The initial condition appears
                 in both approaches defining a realisation of an
                 intrinsic solution (we termed Cognitive Structure) that
                 is holistic, i.e., complete and self-contained. It?s a
                 solution not for one single problem, but for a large
                 class of similar problems. Under this point of view, we
                 would divide any problem solution as two different
                 levels: one for the CS search, and the other to its
                 adaptation to one realization. A general overview is:
                 the information available of the problem feeds GP
                 software, with after some generations obtain the
                 cognitive structure of the problem, or the best
                 available at this moment with minimise the fitness
                 function, i.e., the action for the system. This
                 structure needs to be adapted to the real conditions of
                 the system.",
}

Genetic Programming entries for James Cunha Werner

Citations