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

Re: Re[2]: [GP] prerequisites for using GP

Expand Messages
  • Terence Soule
    Hi, I think if you just go to Google scholar and do a search on: microarrays genetic programming you should turn up a number of papers. Here s a few
    Message 1 of 9 , Mar 31, 2009
    • 0 Attachment
      Hi,

      I think if you just go to Google scholar and do a search on:
      microarrays genetic programming
      you should turn up a number of papers. Here's a few citations:
      @article{hong2006ccb,
      title={{The classification of cancer based on DNA microarray data
      that uses diverse ensemble genetic programming}},
      author={Hong, J.H. and Cho, S.B.},
      journal={Artificial Intelligence in Medicine},
      volume={36},
      number={1},
      pages={43--58},
      year={2006},
      publisher={Elsevier}
      }


      @article{langdon2004gpm,
      title={{Genetic programming for mining DNA chip data from cancer patients}},
      author={Langdon, WB and Buxton, BF},
      journal={Genetic Programming and Evolvable Machines},
      volume={5},
      number={3},
      pages={251--257},
      year={2004},
      publisher={Springer}
      }

      Terry

      On Tue, Mar 31, 2009 at 10:11 AM, Sergey Grechin <hq9000_vst@...> wrote:
      > Hi Terry,
      >
      > Could you plase give a name of the paper or link or something that could
      > help me to find more info on this research?
      >
      > Thanks
      >
      > Sergey
      >
      >> I believe that there have been some successes using GP to analyze
      >> microarray data where the ratio of inputs to training data is similar
      >> although the scale is much larger (1000's to 10,000s of input
      >> variables and only 100's of historical data points). You could look
      >> at the techniques that have been used there. Although if you really
      >> only have K=6 I think you're in trouble.
      >
      >> Terry Soule
      >> Department of Computer Science
      >> University of Idaho
      >> Moscow, ID 83843
      >> tsoule@...
      >
      >> On Mon, Mar 30, 2009 at 11:08 PM, Sergey Grechin <hq9000_vst@...> wrote:
      >>> Hi,
      >>>
      >>> Thanks for opinion.
      >>> Any idea where to find for reasonable approach to analyze such data?
      >>> I have no option to get more observations.
      >>>
      >>> I think I can somehow fight this overfitting by harder constrains on
      >>> complexity of the member of the population.
      >>>
      >>> What do you think?
      >>>
      >>> Sergey
      >>>
      >>>>It sounds like any evolutionary search-based approach will overfit
      >>>>massively to such a small dataset.
      >>>>I would not even attempt this problem with GP.
      >>>
      >>>
      >>>
      >>>
      >>>
      >>> ------------------------------------
      >>>
      >>> Yahoo! Groups Links
      >>>
      >>>
      >>>
      >>>
      >>
      >
      >
      >
      > --
      > Ñ óâàæåíèåì,
      >  Sergey                          mailto:hq9000_vst@...
      >
      >
      >
      > ------------------------------------
      >
      > Yahoo! Groups Links
      >
      >
      >
      >
    • adil raja
      Salam mate, Can u give me a call at my number at 0323 444 9019? Best, Adil Raja ________________________________ From: R. Muhammad Atif Azad
      Message 2 of 9 , Apr 1, 2009
      • 0 Attachment
        Salam mate,
        Can u give me a call at my number at 0323 444 9019?

        Best,
        Adil Raja




        ________________________________
        From: R. Muhammad Atif Azad <atif.azad@...>
        To: genetic_programming@yahoogroups.com
        Sent: Tuesday, March 31, 2009 5:44:31 PM
        Subject: Re: [GP] prerequisites for using GP




        On Tue, 31 Mar 2009, R. Muhammad Atif Azad wrote:

        I agree that your best hope is GP simplifying the functional mapping: it may
        make a large number of inputs redundant as well. One obvious benchmark may be
        (with least squares regression) would be to keep the degree of the polynomial
        at least less than the number of sample points (or else there is no data
        compression achieved). To avoid overfitting in that case, you may also
        consider ridge regression.

        On Mon, 30 Mar 2009, Sergey Grechin wrote:

        >>
        >> Hi,
        >>
        >> Thanks for opinion.
        >> Any idea where to find for reasonable approach to analyze such data?
        >> I have no option to get more observations.
        >>
        >> I think I can somehow fight this overfitting by harder constrains on
        >> complexity of the member of the population.
        >>
        >> What do you think?
        >>
        >> Sergey
        >>
        >> >It sounds like any evolutionary search-based approach will overfit
        >> >massively to such a small dataset.
        >> >I would not even attempt this problem with GP.






        [Non-text portions of this message have been removed]
      Your message has been successfully submitted and would be delivered to recipients shortly.