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

Slow learning

Expand Messages
  • steve@easynn.com
    Deep in the control logic of EasyNN-plus is a diagnostic facility that adds a delay to each cycle so that learning can be slowed down. It was first included so
    Message 1 of 6 , Oct 3, 2003
    • 0 Attachment
      Deep in the control logic of EasyNN-plus is a diagnostic facility that
      adds a delay to each cycle so that learning can be slowed down. It was
      first included so that I could get some trace data during network
      oscillations. That is, I see an oscillation and hit the stop button.
      The idea being that I get the trace at the point of oscillation and
      stop before the buffer overflows.

      Yesterday I was asked about watching the network while learning. Speed
      control could be implemented quite easily if I make the diagnostic
      delay available. My only reservation is that it means yet another
      build and release.

      I think it would only be of use in demonstrating the learning process.

      Can we stand another release?


      Steve
      --
      EasyNN-plus. The easy way to build neural networks.
      Build networks from numeric, text and image files.
      http://www.easynn.com
    • walt
      as long as the feature is optional... ... From: steve@easynn.com [mailto:steve@easynn.com] Sent: Friday, October 03, 2003 1:50 PM To: easynn@yahoogroups.com
      Message 2 of 6 , Oct 3, 2003
      • 0 Attachment
        as long as the feature is optional...
        -----Original Message-----
        From: steve@... [mailto:steve@...]
        Sent: Friday, October 03, 2003 1:50 PM
        To: easynn@yahoogroups.com
        Subject: [easynn] Slow learning


        Deep in the control logic of EasyNN-plus is a diagnostic facility that
        adds a delay to each cycle so that learning can be slowed down. It was
        first included so that I could get some trace data during network
        oscillations. That is, I see an oscillation and hit the stop button.
        The idea being that I get the trace at the point of oscillation and
        stop before the buffer overflows.

        Yesterday I was asked about watching the network while learning. Speed
        control could be implemented quite easily if I make the diagnostic
        delay available. My only reservation is that it means yet another
        build and release.

        I think it would only be of use in demonstrating the learning process.

        Can we stand another release?


        Steve
        --
        EasyNN-plus. The easy way to build neural networks.
        Build networks from numeric, text and image files.
        http://www.easynn.com

        Yahoo! Groups Sponsor
        ADVERTISEMENT




        To unsubscribe from this group, send an email to:
        easynn-unsubscribe@yahoogroups.com



        Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.


        [Non-text portions of this message have been removed]
      • steve@easynn.com
        ... The number of milliseconds to delay each cycle is set in the Controls. It must also be enabled with a checkbox so it s difficult to put on unintentionally.
        Message 3 of 6 , Oct 4, 2003
        • 0 Attachment
          On Fri, 03 Oct 2003 18:40:00 -0400, you wrote:

          >as long as the feature is optional...

          The number of milliseconds to delay each cycle is set in the Controls.
          It must also be enabled with a checkbox so it's difficult to put on
          unintentionally.


          Steve
          --
          EasyNN-plus. The easy way to build neural networks.
          Build networks from numeric, text and image files.
          http://www.easynn.com
        • Cayzer, Steve
          As the culprit responsible for this suggestion, I guess I should respond. What I am interested in is showing a class of beginners (1st year undergrads) how a
          Message 4 of 6 , Oct 6, 2003
          • 0 Attachment
            As the culprit responsible for this suggestion, I guess I should respond.

            What I am interested in is showing a class of beginners (1st year
            undergrads) how a simple ANN learns, visually and without the mathematics.
            EasyNN on, say, XOR seems ideal for the task (it's also nice in that there
            is more than one solution to the problem, but I digress)

            Anyway, Steve suggested that I use that 'stop at cycle' feature to show this
            learning process (a feature I had not noticed in my 30 minute evaluation
            exercise)
            - this workaround works fine for me, so I'm not pushing for the diagnostic
            delay any more.

            Having said that, it sounds a good idea to add into the next release, but
            probably doesn't warrant a release on its own.

            Just my €0.02 ;)

            Steve

            PS Hello everybody! from an EasyNN newbie. Nice software!

            > -----Original Message-----
            > From: steve@... [mailto:steve@...]
            > Sent: 03 October 2003 18:50
            > To: easynn@yahoogroups.com
            > Subject: [easynn] Slow learning
            >
            >
            > Deep in the control logic of EasyNN-plus is a diagnostic
            > facility that adds a delay to each cycle so that learning can
            > be slowed down. It was first included so that I could get
            > some trace data during network oscillations. That is, I see
            > an oscillation and hit the stop button. The idea being that I
            > get the trace at the point of oscillation and stop before the
            > buffer overflows.
            >
            > Yesterday I was asked about watching the network while
            > learning. Speed control could be implemented quite easily if
            > I make the diagnostic delay available. My only reservation is
            > that it means yet another build and release.
            >
            > I think it would only be of use in demonstrating the learning process.
            >
            > Can we stand another release?
            >
            >
            > Steve
            > --
            > EasyNN-plus. The easy way to build neural networks.
            > Build networks from numeric, text and image files.
            http://www.easynn.com

            ------------------------ Yahoo! Groups Sponsor ---------------------~--> Buy
            Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark Printer
            at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
            http://www.c1tracking.com/l.asp?cid=5511
            http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/7brrlB/TM
            ---------------------------------------------------------------------~->

            To unsubscribe from this group, send an email to:
            easynn-unsubscribe@yahoogroups.com



            Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
          • steve@easynn.com
            I had an hour to spare so slow learning is implemented and now available. The 3.0i release has a cycle delay and checkbox in the Controls dialog. To switch on
            Message 5 of 6 , Oct 6, 2003
            • 0 Attachment
              I had an hour to spare so slow learning is implemented and now
              available.

              The 3.0i release has a cycle delay and checkbox in the Controls
              dialog. To switch on slow learning enter a suitable cycle delay and
              check the box. Very simple networks like XOR need a delay of about 500
              milliseconds so they can be seen to be changing as learning
              progresses.

              That is the only change from 3.0h to 3.0i


              Steve
              --
              EasyNN-plus. The easy way to build neural networks.
              Build networks from numeric, text and image files.
              http://www.easynn.com
            • Cayzer, Steve
              Cool! Look forward to trying it out with my students Steve ... http://www.easynn.com ... Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
              Message 6 of 6 , Oct 6, 2003
              • 0 Attachment
                Cool! Look forward to trying it out with my students

                Steve

                > -----Original Message-----
                > From: steve@... [mailto:steve@...]
                > Sent: 06 October 2003 10:06
                > To: easynn@yahoogroups.com
                > Subject: Re: [easynn] Slow learning
                >
                >
                > I had an hour to spare so slow learning is implemented and
                > now available.
                >
                > The 3.0i release has a cycle delay and checkbox in the
                > Controls dialog. To switch on slow learning enter a suitable
                > cycle delay and check the box. Very simple networks like XOR
                > need a delay of about 500 milliseconds so they can be seen to
                > be changing as learning progresses.
                >
                > That is the only change from 3.0h to 3.0i
                >
                >
                > Steve
                > --
                > EasyNN-plus. The easy way to build neural networks.
                > Build networks from numeric, text and image files.
                http://www.easynn.com

                ------------------------ Yahoo! Groups Sponsor ---------------------~--> Buy
                Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark Printer
                at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
                http://www.c1tracking.com/l.asp?cid=5511
                http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/7brrlB/TM
                ---------------------------------------------------------------------~->

                To unsubscribe from this group, send an email to:
                easynn-unsubscribe@yahoogroups.com



                Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
              Your message has been successfully submitted and would be delivered to recipients shortly.