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

Re: [GP] newbie ..i could relate schema and genetic programming

Expand Messages
  • Riccardo Poli
    Hi Manjunath, ... At http://cswww.essex.ac.uk/staff/poli/papers/schema-theory.html you will be able to find a list of about 20 papers on the topic. Journal
    Message 1 of 4 , Nov 19, 2001
    • 0 Attachment
      Hi Manjunath,

      >
      > Im newbie to genetic programming , i took introductory class
      > then i read about schemas i could not relate schemas and gentic
      > programming.
      >
      > i would appreciate help in this regard
      >
      > I would like have online references to read about schemas.
      >

      At

      http://cswww.essex.ac.uk/staff/poli/papers/schema-theory.html

      you will be able to find a list of about 20 papers on the topic.
      Journal papers are generally more suitable for beginners.

      Bill Langdon and I have co-authored a book (Foundations of Genetic
      Programming, Springer, due this month) which includes several
      chapters on schema theory for genetic programming and it shows how it
      relates to the GA theory. See,

      http://www.cs.ucl.ac.uk/staff/W.Langdon/FOGP/

      for more information.


      Riccardo

      --
      ----------------------------------------------------------------------
      Prof Riccardo Poli E-mail: rpoli@...
      Department of Computer Science Room Number: 4B.524
      University of Essex Telephone: (01206) 872338
      Wivenhoe Park Facsimile: (01206) 872788
      Colchester CO4 3SQ, UK http://cswww.essex.ac.uk/staff/poli/
      ----------------------------------------------------------------------
    • Riccardo Poli
      Hi Nina, ... This is true. There is, however, a whole class of GP operators, which I call homologous crossovers, where the position of the genetic material
      Message 2 of 4 , Nov 19, 2001
      • 0 Attachment
        Hi Nina,

        > I too am just learning about GP, and have discovered the debate
        > surrounding GP-Schemata to be very interesting. Essentially defining
        > GP-Schemata is tricky because a there is no inherent semantic to a
        > tree-structure (it's essential "unordered"). ie basic GA crossover
        > preserves the relative position of the bits within the solutions, whereas
        > basic GP crossover allows subtrees to be swapped to a new relative
        > position in the tree...

        This is true. There is, however, a whole class of GP operators, which
        I call homologous crossovers, where the position of the genetic
        material taken from the parents when creating the offspring is
        preserved. Examples of operators in this class include: GP one-point
        crossover (which Bill Langdon and I defined in our GP'97 paper and
        Evol. Comput. 1998 paper) and GP uniform crossover (defined in my GP98
        paper). These are generalisations of the corresponding GA operators.

        >
        > It seems that the bulk of the work in this field is produced by the
        > University of Birmingham (in the UK), primarily by Riccardo Poli.

        Yes, and lots of wonderful people are still there, but I have recently
        moved to the University of Essex for a chair.

        >
        > A great place to start is "A Review of Theoritical and Experimental
        > Results on Schemata in Genetic Programming" by R. Poli and W.B. Langdon.
        > They provide an overview/analysis of some of the different definitions of
        > GP schemata that have been proposed. (Be warned that many papers in
        > this field contain alot of dense mathematical proofs, but this
        > particular paper is reasonably accessible)

        Thank you very much for your interest in our work.

        I just wanted to mention that in 1997/98 we did not have GP schema
        theorems giving exact predictions: we only had lower bounds. It was
        only a couple of years ago that eventually we cracked this hard
        nut. So, if you are interested in this topic, I would suggest you
        considered more recent and extensive work, e.g. my paper in the
        Journal of Genetic Programming and Evolvable Machines vol. 2, n.2,
        2001, or our Foundations of Genetic Programming book.


        Riccardo

        --
        ----------------------------------------------------------------------
        Prof Riccardo Poli E-mail: rpoli@...
        Department of Computer Science Room Number: 4B.524
        University of Essex Telephone: (01206) 872338
        Wivenhoe Park Facsimile: (01206) 872788
        Colchester CO4 3SQ, UK http://cswww.essex.ac.uk/staff/poli/
        ----------------------------------------------------------------------
      Your message has been successfully submitted and would be delivered to recipients shortly.