Loading ...
Sorry, an error occurred while loading the content.
 

IlliGAL New Technical Reports Announcement (November 2001)

Expand Messages
  • Martin Pelikan
    The Illinois Genetic Algorithms Laboratory (IlliGAL) is pleased to announce the publication of the following new technical reports and software. Most IlliGAL
    Message 1 of 4 , Nov 1, 2001
      The Illinois Genetic Algorithms Laboratory (IlliGAL) is pleased to
      announce the publication of the following new technical reports and
      software. Most IlliGAL technical reports, as well as reprints of other
      publications, are available in hardcopy and can be ordered from the
      IlliGAL librarian, (see below for ordering information). The technical
      reports in this announcement are also available electronically on our ftp
      and WWW servers (see the end of this announcement for ftp and WWW access
      instructions).

      --------------------------

      IlliGAL Report No 2001027

      Classifiers that Approximate Functions

      Wilson, S.W.

      Abstract:
      A classifier system, XCSF, is introduced in which the prediction
      estimation mechanism is used to learn approximations to functions. The
      addition of weight vectors to the classifiers allows piecewise-linear
      approximation, where the classifier's prediction is calculated instead of
      being a fixed scalar. Results on functions of up to six dimensions show
      high accuracy. The idea of calculating the prediction leads to the
      concept of a generalized classifier in which the payoff prediction
      approximates the environmental payoff function over a subspace defined by
      the classifier condition and an action restriction specified in the
      classifier, permitting continuous-valued actions.

      --------------------------

      IlliGAL Report No 2001028

      Convergence-time models for the simple genetic algorithm with finite
      population

      Ceroni, A., Pelikan, M., Goldberg, D.E.

      Abstract:
      This paper presents various convergence models for the simple genetic
      algorithm (SGA) in the case of finite population. A piecewise
      convergence-time model is derived using ideas from two existing
      convergence models. The factors affecting the convergence with small
      population size are explained and used to construct a correct model of
      the variance in fitness for the OneMax problem. This knowledge is
      included in the existing asymptotic model to derive the embedded
      convergence-time model. The model is extended to a different environment
      and modified to include an unexpected behavior that makes the SGA
      converge solely by genetic drift.

      --------------------------

      RETRIEVAL/ORDERING:

      The above IlliGAL reports and publications, along with other
      publications and source code, are available electronically via FTP or
      WWW, or as hardcopy directly from us:

      FTP: ftp ftp-illigal.ge.uiuc.edu
      login: anonymous
      password: (your email address)
      cd /pub/papers/IlliGALs (for reports) or
      cd /pub/papers/Publications (for preprints) or
      cd /pub/src (for GA and classifier system source code)
      binary
      get 99022.ps.Z (for example)

      Please look at the README files for explanations of what the file
      names mean. IlliGAL reports are all compressed postscript files.

      WWW: To access the IlliGAL home page, open
      http://www-illigal.ge.uiuc.edu/

      HARDCOPY:

      You can also order hardcopy versions of most IlliGAL publications
      Use the order form in the web or request them directly
      (by IlliGAL number or title) from the IlliGAL librarian:

      Internet: library@... Phone: 217/333-2346
      Fax: 217/244-5705
      Surface mail: IlliGAL Librarian
      Department of General Engineering
      117 Transportation Building
      104 South Mathews Avenue
      Urbana, IL 61801-2996, USA

      When ordering hardcopy, please include your surface mail address!

      Martin Pelikan

      ----------------------------------------------
      Martin Pelikan
      Illinois Genetic Algorithms Laboratory
      University of Illinois at Urbana Champaign
      117 Transportation Building
      104 S. Mathews Avenue, Urbana, IL 61801
      Tel: (217) 333-2346, FAX: (217) 244-5705
      E-mail: pelikan@...
      WWW: http://www-illigal.ge.uiuc.edu/~pelikan/
      ----------------------------------------------
    • Nina Thiessen
      greetings, I m experimenting with lilgp for the first time. Looking at my .gen output file it begins like the following: === GENERATION 0 === total population:
      Message 2 of 4 , Nov 2, 2001
        greetings,

        I'm experimenting with lilgp for the first time. Looking at my .gen output
        file it begins like the following:

        === GENERATION 0 ===
        total population:
        generation:
        mean: nodes: 12.525 (1-63); depth: 3.675 (0-6)
        best: nodes: 18; depth: 5
        worst: nodes: 47; depth: 6
        run: (200 trees)
        mean: nodes: 12.525 (1-63); depth: 3.675 (0-6)
        best: nodes: 18; depth: 5
        worst: nodes: 47; depth: 6
        === GENERATION 1 ===
        total population:
        generation:
        mean: nodes: 9.980 (1-41); depth: 3.590 (0-10)
        best: nodes: 12; depth: 6
        worst: nodes: 15; depth: 6
        run: (400 trees)
        mean: nodes: 11.252 (1-63); depth: 3.632 (0-10)
        best: nodes: 12; depth: 6
        worst: nodes: 47; depth: 6
        === GENERATION 2 ===
        total population:
        generation:
        mean: nodes: 8.385 (1-46); depth: 3.150 (0-9)
        best: nodes: 14; depth: 5
        worst: nodes: 11; depth: 4
        run: (600 trees)
        mean: nodes: 10.297 (1-63); depth: 3.472 (0-10)
        best: nodes: 14; depth: 5

        etc...

        I'm assuming "run: (n trees)" means there are n-many trees in that
        generation's population. When I set pop_size = 200, I expected every
        generation to contain 200 individuals, not for the population size to
        increase by 200 on every iteration! Is there some way to maintain a
        constant population size?

        thanks,
        Nina

        PS. If you happen to know where this is explained in the docs please let
        me know. (I suspect this info is in there, but for some reason I was
        unable to find it.)
      • Terry Van Belle
        Hi Nina, Don t worry, the population is staying constant at 200. What you re looking at is the total run stats, which gives the mean/best/worst for *all*
        Message 3 of 4 , Nov 2, 2001
          Hi Nina,

          Don't worry, the population is staying constant at 200. What you're
          looking at is the total run stats, which gives the mean/best/worst for
          *all* generations up to and including the current generation.

          In other words, in generation 0, lilgp has evaluated 200 trees (the
          initial randomly generated trees). In generation 1, lilgp has evaluated
          400 trees (the initial 200 + the 200 generated by crossover/mutation) to
          get its stats, and so on. Thus, the total number of trees evaluated
          increases by 200 each generation, even though the population size remains
          constant.

          Terry

          On Fri, 2 Nov 2001, Nina Thiessen wrote:

          > greetings,
          >
          > I'm experimenting with lilgp for the first time. Looking at my .gen output
          > file it begins like the following:
          >
          > === GENERATION 0 ===
          > total population:
          > generation:
          > mean: nodes: 12.525 (1-63); depth: 3.675 (0-6)
          > best: nodes: 18; depth: 5
          > worst: nodes: 47; depth: 6
          > run: (200 trees)
          > mean: nodes: 12.525 (1-63); depth: 3.675 (0-6)
          > best: nodes: 18; depth: 5
          > worst: nodes: 47; depth: 6
          > === GENERATION 1 ===
          > total population:
          > generation:
          > mean: nodes: 9.980 (1-41); depth: 3.590 (0-10)
          > best: nodes: 12; depth: 6
          > worst: nodes: 15; depth: 6
          > run: (400 trees)
          > mean: nodes: 11.252 (1-63); depth: 3.632 (0-10)
          > best: nodes: 12; depth: 6
          > worst: nodes: 47; depth: 6
          > === GENERATION 2 ===
          > total population:
          > generation:
          > mean: nodes: 8.385 (1-46); depth: 3.150 (0-9)
          > best: nodes: 14; depth: 5
          > worst: nodes: 11; depth: 4
          > run: (600 trees)
          > mean: nodes: 10.297 (1-63); depth: 3.472 (0-10)
          > best: nodes: 14; depth: 5
          >
          > etc...
          >
          > I'm assuming "run: (n trees)" means there are n-many trees in that
          > generation's population. When I set pop_size = 200, I expected every
          > generation to contain 200 individuals, not for the population size to
          > increase by 200 on every iteration! Is there some way to maintain a
          > constant population size?
          >
          > thanks,
          > Nina
          >
          > PS. If you happen to know where this is explained in the docs please let
          > me know. (I suspect this info is in there, but for some reason I was
          > unable to find it.)
          >
          >
          >
          > To unsubscribe from this group, send an email to:
          > genetic_programming-unsubscribe@yahoogroups.com
          >
          >
          >
          > Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
          >
          >
          >
        • Sean Luke
          A little while ago on this list, someone mentioned in passing that he had done GP using prolog. I m gathering information on existing attempts of this type.
          Message 4 of 4 , Nov 2, 2001
            A little while ago on this list, someone mentioned in passing that he had
            done GP using prolog. I'm gathering information on existing attempts of
            this type. If you've done rule-based GP, I'd be interested in hearing
            about it.

            Sean, who's already got info on l-systems an Pitt-approach rulesystems
          Your message has been successfully submitted and would be delivered to recipients shortly.