• Analog error correction codes using fractals

    From sean.c4s.vn@gmail.com@21:1/5 to All on Wed Aug 26 05:09:24 2015
    I am interest in analog error correction codes at the moment. One way of doing them is with fractals:
    http://arxiv.org/pdf/1105.1561.pdf
    Does anyone have any ideas of better or more configurable fractals that could be used?
    The idea of an error correction code is that it allows you to remove noise from a signal better than averaging it n times, up to a certain maximum signal to noise ratio.
    An application would be with neural nets.
    Sean O'Connor

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Roger Bagula@21:1/5 to sean....@gmail.com on Thu Aug 27 08:23:47 2015
    That depends on the amplitude of the noise:
    doing a fractal encoding of the image allows low amplitude noise
    that doesn't follow the fractal pattern to be eleminated.
    See books by Barnsley and Fisher: http://www.amazon.com/Fractal-Image-Compression-Michael-Barnsley/dp/B005S0TINM/ref=sr_1_7?s=books&ie=UTF8&qid=1440688972&sr=1-7&keywords=michael+F+barnsley
    http://www.amazon.com/Fractal-Image-Compression-Application-Construction/dp/0387942114/ref=sr_1_2?ie=UTF8&qid=1440688877&sr=8-2&keywords=yuval+fisher

    On Wednesday, August 26, 2015 at 5:09:25 AM UTC-7, sean....@gmail.com wrote:
    I am interest in analog error correction codes at the moment. One way of doing them is with fractals:
    http://arxiv.org/pdf/1105.1561.pdf
    Does anyone have any ideas of better or more configurable fractals that could be used?
    The idea of an error correction code is that it allows you to remove noise from a signal better than averaging it n times, up to a certain maximum signal to noise ratio.
    An application would be with neural nets.
    Sean O'Connor

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Roger Bagula@21:1/5 to All on Thu Aug 27 08:29:59 2015
  • From sean.c4s.vn@gmail.com@21:1/5 to All on Thu Aug 27 09:30:52 2015
    I think the idea is slightly different to IFS systems being slightly resistant to noise. You are given a list of real numbers that you want to transmit over a noisy channel. So you add extra numbers to the list that will allow better reconstruction at
    the receiving end. The extra numbers being a fractal walk starting from the numbers in the original list.
    When you have received the data contaminated with noise you have to find a best fit for all the numbers in the list including the extra ones. That is kind of like a lock and key as long as the noise contamination is not too great. You should be able to
    get your original list of numbers back in pretty much exact form.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)