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

RE: [GP] Intelligence necessary for innovation?

Expand Messages
  • Lanre Amos
    Here we go again! Of-course, theistic theory has to be allowed. The atheists who claim evolution ONLY are the problem. It s simple really...teach the kids all
    Message 1 of 18 , Feb 11, 2002
    • 0 Attachment
      Here we go again!

      Of-course, theistic theory has to be allowed. The atheists who claim
      evolution ONLY are the problem.

      It's simple really...teach the kids all that we know and don't know
      until the matter is finally resolved one way or the other.

      Lanre.


      -----Original Message-----
      From: Collin Lynch [mailto:cfl96@...]
      Sent: Monday, February 11, 2002 9:29 AM
      To: genetic_programming@yahoogroups.com
      Subject: [GP] Intelligence necessary for innovation?


      The NY Times (free reg required) is reporting on a challenge to
      evolution
      being mounted in Ohio. The Theory of Intelligent Design holds that
      intelligence (possibly God but not necessarily) is necessary to get the
      process of evolution (or indeed anything) started. They are seeking
      placement alongside evolution in Ohio schools.

      http://www.nytimes.com/2002/02/11/education/11CREA.html?todaysheadlines

      The Intelligent Designers argue that they have the support of
      "Well-credentialed scientists" and that:

      "Science has no statement to make beyond the natural world, ...
      Intelligent design is about how things got started. Evolution is
      about how they change across time."

      The chief opponent of the change is, ironically, a creationist who
      argues
      that this is a veiled attempt to insert theism into schools. Opponents
      also argue that this is, by no means a rigorously tested theory.

      As people who use their intelligence to design, craft, and monitor
      evolutionary processes what do you all think? Are any if not all of the
      arguments credible, or are they walking down the same road as Kansas
      with
      with a different flag?

      Collin Lynch.
      Research Programmer, University of Pittsburgh.





      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/
    • Gordon D. Pusch
      ... The problem with such a free-wheeling ``live and let live approach is that the Public Educational system is fundamentally AUTHORITARIAN down to its very
      Message 2 of 18 , Feb 11, 2002
      • 0 Attachment
        Lanre Amos <lanre@...> writes:

        > Here we go again!
        >
        > Of-course, theistic theory has to be allowed. The atheists who claim
        > evolution ONLY are the problem.
        >
        > It's simple really...teach the kids all that we know and don't know
        > until the matter is finally resolved one way or the other.

        The problem with such a free-wheeling ``live and let live'' approach
        is that the Public Educational system is fundamentally AUTHORITARIAN
        down to its very roots as an arm of the Federal Gov't and an agent of
        Social Engineering, so the idea of teaching anything but ``official truth''
        or allowing students to learn to be critical and decide for themselves
        is utterly anathema to the bureaucracy that funds and runs the public
        schools.

        You must realize that the primary purpose of public schools for nearly
        a century now has been INDOCTRINATION, not ``education.'' This can be
        shown no more clearly than in the own words of John Dewey, the ``Father
        of Modern Education'' and an ardent socialist:

        ``...The process of merely acquiring knowledge for its own sake is
        so intensely personal an experience that it can easily lead to
        selfishness. It therefore serves _NO USEFUL SOCIAL PURPOSE_...''

        [my emphasis]. Dewey made it absolutely crystal clear in his writings
        that the goal of ``education'' should be _SOCIALIZATION_, not learning.
        He wanted to engineer the ``New Soviet Man'' who would always sacrifice
        himself on the altar of ``The Needs of Society'' --- so producing critical
        thinkers was the very =LAST= thing on his mind! This philosophy is still
        at the core of nearly every school of education in the US, and we are now
        reaping a harvest of more than 60% functional illiteracy because of it.
        The current incarnations of this philosophy are ``whole language education''
        (which among other things encourages ``creative spelling and grammar''),
        and ``outcomes-based education,'' which holds that the most important
        function of the public school is to ``improve the self-esteem of students''
        and ``eliminate competitiveness and acting out'' by (among other things)
        eliminating performance-based hierarchical letter-grade in favor of
        a set of seven ``positive and affirming'' adjectives. In other words,
        ``bad students'' will be eliminated by eliminating the concept of
        ``passing and failing'' (thereby defining them out of existence),
        and there will be no more functional illiteracy because there will be
        no more standards --- the bar will be set at ground level.

        Welcome to the age of C.M. Kornbluth's ``Marching Morons''... :-(


        -- Gordon D. Pusch

        perl -e '$_ = "gdpusch\@...\n"; s/NO\.//; s/SPAM\.//; print;'
      • Martin Sewell
        ... Why? Belief in the existence of a god or gods is a violation of Occam s razor. ... No they re not. As science has advanced, atheism is now the only
        Message 3 of 18 , Feb 11, 2002
        • 0 Attachment
          At 09:49 11/02/2002 -0800, Lanre Amos wrote:
          >Here we go again!
          >
          >Of-course, theistic theory has to be allowed.

          Why? Belief in the existence of a god or gods is a violation of Occam's razor.

          >The atheists who claim evolution ONLY are the problem.

          No they're not. As science has advanced, atheism is now the only sensible
          position to adopt.

          >It's simple really...teach the kids all that we know and don't know until
          >the matter is finally resolved one way or the other.

          Modern science is ultimately based on Bayesian statistics, which tells us
          that lack of evidence for a proposition is evidence for its negation. This
          gives us grounds for not teaching certain things.

          Regards

          Martin
        • Lanre Amos
          I have no problems with your beliefs. I ll even go so far as to agree that lack of evidence for a proposition is evidence for its negation. We are talking
          Message 4 of 18 , Feb 11, 2002
          • 0 Attachment
            I have no problems with your beliefs. I'll even go so far as to agree
            that lack of evidence for a proposition is evidence for its negation.

            We are talking evidence...not proof. From where I stand, not only is
            there evidence of Intelligent Design, there is super abundant evidence
            for Intelligent Evolution.

            It's only honest to not try to impose one's religious beliefs on others
            under the guise of science.

            Lanre.





            -----Original Message-----
            From: Martin Sewell [mailto:M.Sewell@...]
            Sent: Monday, February 11, 2002 4:34 PM
            To: genetic_programming@yahoogroups.com
            Subject: RE: [GP] Intelligence necessary for innovation?


            At 09:49 11/02/2002 -0800, Lanre Amos wrote:
            >Here we go again!
            >
            >Of-course, theistic theory has to be allowed.

            Why? Belief in the existence of a god or gods is a violation of Occam's
            razor.

            >The atheists who claim evolution ONLY are the problem.

            No they're not. As science has advanced, atheism is now the only
            sensible
            position to adopt.

            >It's simple really...teach the kids all that we know and don't know
            until
            >the matter is finally resolved one way or the other.

            Modern science is ultimately based on Bayesian statistics, which tells
            us
            that lack of evidence for a proposition is evidence for its negation.
            This
            gives us grounds for not teaching certain things.

            Regards

            Martin



            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/
          • Martin Sewell
            ... What evidence do you believe exists? Intelligent Design isn t a scientific theory, it is metaphysical. ... Agreed. Regards Martin
            Message 5 of 18 , Feb 11, 2002
            • 0 Attachment
              At 17:17 11/02/2002 -0800, Lanre Amos wrote:
              >I have no problems with your beliefs. I'll even go so far as to agree that
              >lack of evidence for a proposition is evidence for its negation.
              >
              >We are talking evidence...not proof. From where I stand, not only is there
              >evidence of Intelligent Design, there is super abundant evidence for
              >Intelligent Evolution.

              What evidence do you believe exists? Intelligent Design isn't a scientific
              theory, it is metaphysical.

              >It's only honest to not try to impose one's religious beliefs on others
              >under the guise of science.

              Agreed.

              Regards

              Martin
            • Terry Van Belle
              I would humbly submit that this conversation is off-topic for this group, having nothing to do with Genetic Programming, and would ask the two of you to take
              Message 6 of 18 , Feb 11, 2002
              • 0 Attachment
                I would humbly submit that this conversation is off-topic for this group,
                having nothing to do with Genetic Programming, and would ask the two of
                you to take it off-line please.

                Terry

                On Tue, 12 Feb 2002, Martin Sewell wrote:

                > At 17:17 11/02/2002 -0800, Lanre Amos wrote:
                > >I have no problems with your beliefs. I'll even go so far as to agree that
                > >lack of evidence for a proposition is evidence for its negation.
                > >
                > >We are talking evidence...not proof. From where I stand, not only is there
                > >evidence of Intelligent Design, there is super abundant evidence for
                > >Intelligent Evolution.
                >
                > What evidence do you believe exists? Intelligent Design isn't a scientific
                > theory, it is metaphysical.
                >
                > >It's only honest to not try to impose one's religious beliefs on others
                > >under the guise of science.
                >
                > Agreed.
                >
                > Regards
                >
                > Martin
                >
                >
                >
                > 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/
                >
                >
                >
              • Lanre Amos
                Terry, I think it might be relevant actually. ADF s and other enhancements might be considered small attempts at Intelligent Evolution. As someone just waiting
                Message 7 of 18 , Feb 11, 2002
                • 0 Attachment
                  Terry, I think it might be relevant actually.

                  ADF's and other enhancements might be considered small attempts at
                  Intelligent Evolution.

                  As someone just waiting to apply the results of GP research, I've always
                  wondered if anybody has tried to introduce explicit gradient information
                  into GP.

                  Or is that heretical?

                  Lanre.



                  -----Original Message-----
                  From: Terry Van Belle [mailto:vanbelle@...]
                  Sent: Monday, February 11, 2002 6:41 PM
                  To: genetic_programming@yahoogroups.com
                  Subject: RE: [GP] Intelligence necessary for innovation?


                  I would humbly submit that this conversation is off-topic for this
                  group,
                  having nothing to do with Genetic Programming, and would ask the two of
                  you to take it off-line please.

                  Terry

                  On Tue, 12 Feb 2002, Martin Sewell wrote:

                  > At 17:17 11/02/2002 -0800, Lanre Amos wrote:
                  > >I have no problems with your beliefs. I'll even go so far as to agree
                  that
                  > >lack of evidence for a proposition is evidence for its negation.
                  > >
                  > >We are talking evidence...not proof. From where I stand, not only is
                  there
                  > >evidence of Intelligent Design, there is super abundant evidence for
                  > >Intelligent Evolution.
                  >
                  > What evidence do you believe exists? Intelligent Design isn't a
                  scientific
                  > theory, it is metaphysical.
                  >
                  > >It's only honest to not try to impose one's religious beliefs on
                  others
                  > >under the guise of science.
                  >
                  > Agreed.
                  >
                  > Regards
                  >
                  > Martin
                  >
                  >
                  >
                  > 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/
                  >
                  >
                  >



                  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/
                • Martin Sewell
                  ... Intelligent Design (ID) refers to the theory that intelligent causes are responsible for the origin of the universe and of life in all its diversity.
                  Message 8 of 18 , Feb 12, 2002
                  • 0 Attachment
                    At 19:41 11/02/2002 -0700, Terry Van Belle wrote:
                    >I would humbly submit that this conversation is off-topic for this group,
                    >having nothing to do with Genetic Programming, and would ask the two of
                    >you to take it off-line please.

                    Intelligent Design (ID) refers to the theory that intelligent causes are
                    responsible for the origin of the universe and of life in all its diversity.

                    Michael Behe, an Associate Professor of Biochemistry at Lehigh University,
                    is an advocate of ID. His argument hinges upon the notion of "irreducibly
                    complex systems," systems that could not function if they were missing just
                    one of their many parts. "Irreducibly complex systems ... cannot evolve in
                    a Darwinian fashion," he says, because natural selection works on small
                    mutations in just one component at a time.

                    The arguments, therefore, may be of interest to the evolutionary
                    computation community.

                    Regards

                    Martin
                  • yseeley
                    Irreducibly complex systems ... cannot evolve in ... small ... In which case, the following link may be of interest:
                    Message 9 of 18 , Feb 12, 2002
                    • 0 Attachment
                      "Irreducibly complex systems ... cannot evolve in
                      > a Darwinian fashion," he says, because natural selection works on
                      small
                      > mutations in just one component at a time.
                      >
                      > The arguments, therefore, may be of interest to the evolutionary
                      > computation community.

                      In which case, the following link may be of interest:
                      http://ucsdnews.ucsd.edu/newsrel/science/mchox.htm

                      FIRST GENETIC EVIDENCE UNCOVERED OF HOW MAJOR CHANGES IN BODY SHAPES
                      OCCURRED DURING EARLY ANIMAL EVOLUTION

                      -Yonik
                    • Sean Luke
                      [My apologies, but since irreducible complexity was proposed in a GP context, I think it s worthwhile examining why the theorem, as far as I understand it, is
                      Message 10 of 18 , Feb 12, 2002
                      • 0 Attachment
                        [My apologies, but since irreducible complexity was proposed in a GP
                        context, I think it's worthwhile examining why the theorem, as far as I
                        understand it, is wrong.]

                        On Tuesday, February 12, 2002, at 07:10 AM, Martin Sewell wrote:

                        > Michael Behe, an Associate Professor of Biochemistry at Lehigh
                        > University,
                        > is an advocate of ID.

                        Michael Behe is famous for a broken theorem known as irreducible
                        complexity, which is widely touted by ID proponents unaware that it's
                        mathematical gibberish. The theorem is dressed up in various ways but
                        it's actually quite simple. Behe stipulates that there is a minimum
                        level of complexity of an organism in order to survive. That is, let's
                        say the organism consists of modules A and B. If you take away either
                        module, then the organism dies because A cannot survive on its own and B
                        cannot survive on its own. How then, Behe asks, did this organism
                        evolve from these two components? It couldn't have, thus it had to be
                        intelligently created in its AB form.

                        Since Behe is stating that it is *impossible* to arrive at AB via
                        evolution, he's making a universal statement: so all you need is a
                        counterexample. Here it is. Behe relies on two assumptions, both of
                        which are false:

                        1. Species survival is a yes/no proposition.
                        2. One can only add modules, not delete or change them.

                        Imagine if an organism can survive, reasonably well, with module C.
                        Adding A or B to C confers an advantage. Adding both A and B to C is
                        even better. Finally, A and B alone is better still. Thus you have:

                        C -> AB -> ABC -> AB

                        Nothing to it.

                        Sean
                      • Jim Wyman
                        GPers might well be attracted to discussions of Behe s irreducible complexity and fellow ID proponent William Dembski s theory of information. Thorough
                        Message 11 of 18 , Feb 12, 2002
                        • 0 Attachment
                          GPers might well be attracted to discussions of Behe's "irreducible
                          complexity" and fellow ID proponent William Dembski's theory of
                          information. Thorough treatment of both can be found in Chapter Four,
                          "Messages from Heaven", of Victor Stenger's upcoming book 'Has
                          Science Found God?: The Latest Results in the Search for Purpose in
                          the Universe', to be published in October, 2002. Chapter Four and two
                          other sample chapters are available in .pdf form at:
                          http://spot.colorado.edu/~vstenger/god.html.

                          On that page is a summary of the book, which is copyrighted, and,
                          although several sample chapters are available, they are still subject
                          to change and correction.

                          Jim Wyman


                          >[My apologies, but since irreducible complexity was proposed in a GP
                          >context, I think it's worthwhile examining why the theorem, as far as I
                          >understand it, is wrong.]
                          >
                          >On Tuesday, February 12, 2002, at 07:10 AM, Martin Sewell wrote:
                          >
                          >> Michael Behe, an Associate Professor of Biochemistry at Lehigh
                          >> University,
                          >> is an advocate of ID.
                          >
                          >Michael Behe is famous for a broken theorem known as irreducible
                          >complexity, which is widely touted by ID proponents unaware that it's
                          >mathematical gibberish. The theorem is dressed up in various ways but
                          >it's actually quite simple. Behe stipulates that there is a minimum
                          >level of complexity of an organism in order to survive. That is, let's
                          >say the organism consists of modules A and B. If you take away either
                          >module, then the organism dies because A cannot survive on its own and B
                          >cannot survive on its own. How then, Behe asks, did this organism
                          >evolve from these two components? It couldn't have, thus it had to be
                          >intelligently created in its AB form.
                          >
                          >Since Behe is stating that it is *impossible* to arrive at AB via
                          >evolution, he's making a universal statement: so all you need is a
                          >counterexample. Here it is. Behe relies on two assumptions, both of
                          >which are false:
                          >
                          >1. Species survival is a yes/no proposition.
                          >2. One can only add modules, not delete or change them.
                          >
                          >Imagine if an organism can survive, reasonably well, with module C.
                          >Adding A or B to C confers an advantage. Adding both A and B to C is
                          >even better. Finally, A and B alone is better still. Thus you have:
                          >
                          >C -> AB -> ABC -> AB
                          >
                          >Nothing to it.
                          >
                          >Sean
                        • Lanre Amos
                          Could it be that the underlying dynamics of scientific research has now entered a local minimum with a basin of attraction so powerful that the global optimum
                          Message 12 of 18 , Feb 13, 2002
                          • 0 Attachment
                            Could it be that the underlying dynamics of scientific research has now
                            entered a local minimum with a basin of attraction so powerful that the
                            global optimum is now out of the reach of the scientific illuminati?

                            The evidence for Intelligent Design in macro politics... the alignment
                            of nations, religions and ideologies... is so clear, it's hard to miss.

                            Glad to be an engineer.

                            Lanre.



                            -----Original Message-----
                            From: Sean Luke [mailto:sean@...]
                            Sent: Tuesday, February 12, 2002 12:29 PM
                            To: genetic_programming@yahoogroups.com
                            Subject: [GP] [Offtopic] Intelligence necessary for innovation?


                            Imagine if an organism can survive, reasonably well, with module C.
                            Adding A or B to C confers an advantage. Adding both A and B to C is
                            even better. Finally, A and B alone is better still. Thus you have:

                            C -> AB -> ABC -> AB

                            Nothing to it.

                            Sean
                          • Tom Osborn
                            ... [Apologies in advance to the general reader. Attempts to amuse have been included to encourage readers to the end. I will not contribute to this thread
                            Message 13 of 18 , Feb 13, 2002
                            • 0 Attachment
                              Lanre Amos chimed in again, for the umpteenth time:
                              >
                              > Could it be that the underlying dynamics of scientific
                              > research has now entered a local minimum with a
                              > basin of attraction so powerful that the global optimum
                              > is now out of the reach of the scientific illuminati? ...

                              [Apologies in advance to the general reader. Attempts to amuse
                              have been included to encourage readers to the end. I will
                              not contribute to this "thread" again].

                              One would recommend to Lanre to set up his own mailing
                              list for non-scientific non-illuminati and people they can
                              impress. This topic re-surfaces time and again in GP_list
                              without any value.

                              My philosophical and mathematical meta-contribution would be
                              that "God" (as an explanation) is either an emergent (cultural)
                              artefact, stemming from initial gullibility and a psychological
                              deficiency (such as a need for prime causes), OR, "God" (as
                              an explanation) is a hidden variable (from the uncountably infinite
                              set of possible hidden variables - and, yes, I know that set should
                              be countable, but there are dual explanators when you have to
                              rely on finite information).

                              In either case, GP is unlikely to resolve the question.

                              The question of intelligent design is a meta-question of some
                              interest. The meta-answer(s) will have to resolve the differences
                              between pragmatic and intentional (intelligent) acts. The latter
                              begs the (meta-)question of hidden variables, while the former
                              begs the question of cultural forming of pragmatic valuation.

                              My political contribution is to question the forms of self-interest
                              which Lanre is serving, and suggest he stop wasting other
                              people's time. If there are errors in majority opinion (which is
                              inevitably the case), his attempts to help as sadly astray.

                              As an observation, I would say I am sick of second rate wankers
                              (apologies for the term, but in this country it's the most apt term)
                              who use the latest scientific language to impress their arguments
                              by *sounding* scientific and "forefront". [A survey of the past 60
                              years of alternative medicine reveals similar use of language].

                              Lastly, I recognise the importance of alternative opinions and
                              thought. just as mutation can be a "saviour" in GP, so are
                              alternative ideas. MOST mutations are deleterious, but without
                              them, escape from local maxes is hindered. In the case of
                              Lanre's apparent contribution, interpreted as an alternative
                              "meme/gene", it's already in the trash can!

                              [Apologies flag off].

                              Please let's stay on topic most of the time. Let's keep speculations
                              and other challenges original and novel.

                              Cheers, Tom.

                              Dr Tom Osborn
                              Director of Modelling
                              The NTF Group
                              http://www.ntf.com.au/
                              Decision Support Consultants
                              Level 7, 1 York Street
                              SYDNEY NSW 2000
                              AUSTRALIA
                              ph: +61 2 9252 0600
                              fax: +61 2 9251 9894
                            • Terence Soule
                              Hi, Does anyone know of any research on operator set choice in GP? I.e. guildlines for which operators to pick for a particular problem or type of problem.
                              Message 14 of 18 , Feb 14, 2002
                              • 0 Attachment
                                Hi,

                                Does anyone know of any research on operator set choice in GP? I.e.
                                guildlines for which operators to pick for a particular problem or type of
                                problem. I've done a search at the standard places (on-line bibliogrpahy
                                and citation index) but haven't found anything. I may just be using the
                                wrong keywords.

                                Thanks for any assistance,
                                Terry Soule
                                tsoule@...
                              • Lanre Amos
                                ... Touche. Mission accomplished. Lanre.
                                Message 15 of 18 , Feb 14, 2002
                                • 0 Attachment
                                  Tom Osborn wrote:
                                  >Lastly, I recognise the importance of alternative opinions and
                                  >thought. just as mutation can be a "saviour" in GP, so are
                                  >alternative ideas. MOST mutations are deleterious, but without
                                  >them, escape from local maxes is hindered. In the case of
                                  >Lanre's apparent contribution, interpreted as an alternative
                                  >"meme/gene", it's already in the trash can!

                                  Touche. Mission accomplished.


                                  Lanre.
                                • Riccardo Poli
                                  Hi Terry, ... As you know there are empirical comparisons of different operators and representations on specific problems and people (myself included) have
                                  Message 16 of 18 , Feb 15, 2002
                                  • 0 Attachment
                                    Hi Terry,

                                    You wrote:
                                    >
                                    > Does anyone know of any research on operator set choice in GP? I.e.
                                    > guildlines for which operators to pick for a particular problem or type of
                                    > problem. I've done a search at the standard places (on-line bibliogrpahy
                                    > and citation index) but haven't found anything.
                                    >

                                    As you know there are empirical comparisons of different operators and
                                    representations on specific problems and people (myself included) have
                                    tried to draw general conclusions (maybe occasionally also qualitative
                                    guidelines) from the observed behaviour. For example, many papers of
                                    this type have been produced a few years ago when the
                                    crossover-is-better-than-mutation
                                    vs. mutation-is-better-than-crossover argument was
                                    fashionable. However, very often it is very easy to find
                                    counter-examples to these general conclusions/guidelines. I suspect it
                                    is not this type of guidelines you are after.

                                    Over time, maybe as a result of the sinking in of the no-free-lunch
                                    results, most people have kind of stopped trying to draw too general
                                    conclusions. More recently it has become clear that you can beat NFL
                                    only if you know something about your problem (or class of problems)
                                    and you have a way of matching that knowledge against knowledge on the
                                    behaviour of different algorithms (by algorithm here I mean: search
                                    algorithm + representation + operators + parameter settings).

                                    However, although it is crucially important to understand what type of
                                    knowledge on problems and algorithms we need in order to perform this
                                    match, I think we have not made much progress. At present we cannot do
                                    much more than apply an algorithm (in the broad sense mentioned above)
                                    to a problem and observe the results. Then, if the results are
                                    statistically significant, we can be sure only that if one used the
                                    same algorithm on the same problem one would get similar
                                    results.

                                    This information is useful. For example, this allows us to compare
                                    different algorithms on the same problem. However, ideally one would
                                    want much more usable results, e.g. guidelines like "if your problem
                                    has features A, B, C choose algorithm X operators Y, Z, W
                                    representation R etc.", where the presence/absence of features A, B, C
                                    can easily, but rigorously, be identified. I think it is this type of
                                    guidelines you refer to in your message.

                                    If so, I am not surprised that you failed to find much literature on
                                    the subject. Very often experience with using GP systems for years
                                    help researchers get representation, operators, parameters etc. right
                                    within a few design iterations. However, there seem to be no easy to
                                    convert this expertise into reliable, general, publishable guidelines.

                                    I believe one way to make progress in this direction is to define and
                                    study tunable mathematical models of the behaviour of
                                    algorithm-problem pairs. By this a mean models which describe the
                                    expected behaviour of an algorithm applied to a specific problem as a
                                    function of: a) the parameters of the algorithm (which could include:
                                    features of the representation, choice and probability of application
                                    of different operators, and so on) and b) some parameters describing
                                    the problem (imagine that the problem belongs to a specific class, and
                                    that you can obtain different instances form that class by modifying a
                                    set of problem parameters).

                                    Nic McPhee and I have started to do exactly this by using and
                                    combining the GP schema equations for three different operators with
                                    the objective of finding optimum combinations of operators (and their
                                    optimum probabilities of application) for three or four different
                                    algorithm performance measures. (Note these should not be confused
                                    with fitness measures: by algorithm performance measures I mean
                                    rigorous mathematical ways of indicating what the objective of running
                                    the algorithm is, e.g. maximising avg fitness, maximising best
                                    fitness, maximising fitness while not growing in size, etc.).

                                    Some results of this work are described in a technical report entitled
                                    "Using schema theory to explore interactions of multiple operators"
                                    available at

                                    http://cswww.essex.ac.uk/technical-reports/2002/tr-csm365.pdf

                                    I append the abstract at the end of this message. Note this paper
                                    describes and iterates a tunable model of the algorithm-problem pair
                                    for a very specific problem and a linear representation. The approach
                                    is probably not going to applicable to all problems, but it gives me
                                    some hope that we can start cracking the problem of matching
                                    algorithms to problems.


                                    Riccardo

                                    ======================================================================
                                    ABSTRACT: In the last two years the schema theory for Genetic
                                    Programming (GP) has been applied to the problem of understanding the
                                    length biases of a variety of crossover and mutation operators on
                                    variable length linear structures. In these initial papers, operators
                                    were studied in isolation. In practice, however, they are typically
                                    used in various combinations, and in this paper we present the first
                                    schema theory analysis of the complex interactions of multiple
                                    operators. In particular we apply the schema theory to the use of
                                    standard subtree crossover, full mutation, and grow mutation (in
                                    varying proportions) to variable length linear structures in the
                                    one-then-zeros problem. We then show how the results can be used to
                                    guide choices about the relative proportion of these operators in
                                    order to achieve certain structural goals during a run.
                                  • Pete Martin
                                    Riccardo, I realise you are only back for a few days, but if you could possibly take a look at the conclusions chapter for me I would be most grateful. If this
                                    Message 17 of 18 , Jun 30, 2002
                                    • 0 Attachment
                                      Riccardo,

                                      I realise you are only back for a few days, but if you could possibly take a
                                      look at the conclusions chapter for me I would be most grateful. If this is
                                      OK, then I hope to be able to print and bind my thesis ready for submission
                                      by the middle of July, after GECCO.

                                      Regards
                                      Pete
                                    • Pete Martin
                                      OOPS! Sorry that was meant to be private mail. Feel free to delete it!. Pete
                                      Message 18 of 18 , Jun 30, 2002
                                      • 0 Attachment
                                        OOPS!

                                        Sorry that was meant to be private mail. Feel free to delete it!.

                                        Pete
                                      Your message has been successfully submitted and would be delivered to recipients shortly.