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

Re: [GP] using CA as an environment for the evolution of predation/defensemechanisms

Expand Messages
  • Natalio Krasnogor
    Dear Tim and Raphael, I m a little bit late with my input but I thought it might anyway be of interest. A few years ago me an colleagues evolved cellular
    Message 1 of 1 , May 3, 2002
      Dear Tim and Raphael,

      I'm a little bit late with my input but I thought it might anyway be of
      A few years ago me an colleagues evolved cellular automata rules where
      the aim
      of the CA was to properly fold a protein.
      A few references that can be of interest to you are:

      * Encoding and Crossover Mismatch in a Molecular Design Problem ,
      Artificial Intelligence in Design '98 (AID98), Esteban de la Canal &
      Krasnogor & Daniel H. Marcos & David Pelta & Walter A. Risi, 1998.

      *Protein Structure Prediction as a Complex Adaptive System , Frontiers
      in Evolutionary Algorithms (FEA98), Natalio Krasnogor & Daniel H. Marcos
      David Pelta & Walter A. Risi, 1998.

      which can be downloaded from my web page:

      The results in these short papers are extended in my colleague, David
      Pelta, thesis that can be found in:

      www.ugr.es/~dpelta/papers.html , following the link "my degree thesis"

      I want to mention though that many interesting, both theoretical and
      issues on the evolution of CA that correctly fold proteins remain open
      (I'll be happy to collaborate with anybody on these lines), and as such
      this topic represents a really challenging problem for GP.


      Tim Taylor wrote:
      > Hi Raphael,
      > I have recently started playing about with a fairly
      > similar system. At present I have been trying to develop
      > a basic understanding of how dynamical systems (and
      > in particular, cellular automata) can be controlled
      > through the use of evolved boundary conditions (e.g.
      > initial configuration of the CA). I am currently using
      > a particular (static) target configuration in the fitness
      > function of a genetic algorithm and trying to evolve
      > genomes which can produce that target (or close to it)
      > by setting a limited number of cell states in the initial
      > configuration. I don't have any publications on this
      > yet, but I am currently preparing a paper to submit
      > to the ALIFE 8 conference. This will focus on the
      > differences in performance of the system under different
      > categories of CA dynamics, as characterised by
      > Langton's lambda parameter.
      > I will be extending the system to look at coevolution
      > in the near future - I think this is much more exciting
      > that the static target configurations that I have
      > currently been using. Indeed, coevolution circumvents
      > some of the problems I have had with these initial
      > studies, e.g. it is, in general, undecidable (I think)
      > whether a particular target configuration is actually
      > achievable for a particular CA transition function or
      > whether it can exist only as a Garden of Eden
      > configuration.
      > I am not aware of anyone else who has been doing this
      > sort of thing. Sure, lots of people have tried evolving
      > a CA's transition function, but that's not quite the
      > same. I would certainly encourage you to go ahead
      > with your study - I think the basic idea that you
      > outlined sounds good. Let me know how you get on!
      > Best wishes,
      > Tim
      > Tim Taylor_____________________________________|office: Room C6
      > Research Fellow, Mobile Robots Group, IPAB |tel:0131-651-1740
      > mail address: Division of Informatics, |fax:0131-650-6899
      > University of Edinburgh, 5 Forrest Hill, Edinburgh, EH1 2QL, U.K.
      > http://www.dai.ed.ac.uk/homes/timt/ mailto:tim.taylor@...
      > On Mon, 29 Apr 2002, Raphael Crawford-Marks wrote:
      > > Howdy GPers,
      > >
      > > This question is a bit more a-lifey, but I'll pose it here since I'm not
      > > subscribed to any A-Life lists at the moment. I'm in the beginning stages
      > > of outlining a system that would, among other things, allow individuals to
      > > evolve predation and defense mechanisms. All throughout nature there are
      > > examples of innovative defense mechanisms that have emerged as defense
      > > against predation (one example I'm familiar with is the California Rough
      > > Skinned newt, which secretes a slightly toxic substance that is very
      > > irritating to mucous membranes - predators tend to spit them out
      > > immediately). Further, some predators have evolved counter-measures to
      > > these defenses. And so on.
      > >
      > > In order to adequately model this phenomenon, the simulation needs to
      > > provide an environment in which this cycle of adaptation and re-adaptation
      > > can occur. My idea was this: Each creature evolves an initial
      > > configuration for some nxn Life gameboard. When one creature preys upon
      > > another, their respective gameboards are joined at the edges, and the rules
      > > of Conway's Life is run for some m generations. When complete, whichever
      > > creature has the most active cells on it's gameboard is the winner (if
      > > predator, the predator eats the prey; if prey, the predator is repelled).
      > > My thinking is that creatures would evolve combinations of glider guns and
      > > eaters in an attempt to maintain a high active cell count on their side
      > > while disrupting the cells on the opposing creatures board.
      > >
      > > Can anyone point me to some related work? Can anyone think of any pitfalls
      > > that might occur with this system? Any other thoughts?
      > >
      > > Thanks,
      > >
      > > Raphael
      > >
      > > --------------------------------------------------------------
      > > Raphael Crawford-Marks
      > > rpc01@... http://hamp.hampshire.edu/~rpc01
      > > Home: 413-559-4614 Cell: 415-596-3500
      > > --------------------------------------------------------------
      > 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/

      NATALIO KRASNOGOR, Ph.D. Computational Biophysics and Chemistry
      Postdoctoral Research Associate School of Chemistry


      Automatic Scheduling and Planning
      (ASAP) group
      School of Computer Sciences and IT

      University Park
      University of Nottingham
      Tel.: +44 - 0115 - 9513477 Nottingham, NG7 2RD
      Fax.: +44 - 0115 - 9513562 United Kingdom

      URL: http://dirac.chem.nott.ac.uk/~natk/Public/index.html

      e-mail: Natalio.Krasnogor@...
    Your message has been successfully submitted and would be delivered to recipients shortly.