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

Back prop learning implementation

Expand Messages
  • tzmtn
    Hi, I ve tried to implement the back prop learning function from page 734 and ran it on the following examples: double[][] inputs = { {0.547361, -2.04845,
    Message 1 of 3 , Apr 16, 2012
    • 0 Attachment
      Hi,
      I've tried to implement the back prop learning function from page 734
      and ran it on the following examples:

      double[][] inputs = {
      {0.547361, -2.04845, -2.71647},
      {1.9708, -1.16141, -0.0485735},
      {-3.18799, 2.97068, 0.26499},
      {-0.498425, -1.04703, 1.61744}
      };

      double[][] outputs = {
      {1.0, 0.0, 0.0, 0.0},
      {0.0, 1.0, 0.0, 0.0},
      {0.0, 0.0, 1.0, 0.0},
      {0.0, 0.0, 0.0, 1.0}
      };

      the problem is my network is unable to fully adapt itself to the data.
      after 10000 iterations of the main loop I still have an error sum of around 6 units. The error doesn't converge to zero. Am I doing something wrong? I'd be glad to share my Java code but I'm new in this group and have no idea how should I post the code.
      Cheers,
      Matan.
    • wrwerwerqwe1e3
      Your training set is fine, you may be missing something in your code. You might find http://pastebin.com/ useful for sharing your code. Greetings, Jorge
      Message 2 of 3 , Apr 17, 2012
      • 0 Attachment
        Your training set is fine, you may be missing something in your code.

        You might find http://pastebin.com/ useful for sharing your code.

        Greetings,
        Jorge

        --- In aima-talk@yahoogroups.com, "tzmtn" <tzmtnh@...> wrote:
        >
        > Hi,
        > I've tried to implement the back prop learning function from page 734
        > and ran it on the following examples:
        >
        > double[][] inputs = {
        > {0.547361, -2.04845, -2.71647},
        > {1.9708, -1.16141, -0.0485735},
        > {-3.18799, 2.97068, 0.26499},
        > {-0.498425, -1.04703, 1.61744}
        > };
        >
        > double[][] outputs = {
        > {1.0, 0.0, 0.0, 0.0},
        > {0.0, 1.0, 0.0, 0.0},
        > {0.0, 0.0, 1.0, 0.0},
        > {0.0, 0.0, 0.0, 1.0}
        > };
        >
        > the problem is my network is unable to fully adapt itself to the data.
        > after 10000 iterations of the main loop I still have an error sum of around 6 units. The error doesn't converge to zero. Am I doing something wrong? I'd be glad to share my Java code but I'm new in this group and have no idea how should I post the code.
        > Cheers,
        > Matan.
        >
      • Matan Halberstadt
        Hi Jorge, Thanks for the reply! Here is my code: http://pastebin.com/SNH6ADVZ As I ve mentioned earlier I have implemented the pseudocode from page 734 in
        Message 3 of 3 , Apr 18, 2012
        • 0 Attachment
          Hi   Jorge,
          Thanks for the reply! 

          Here is my code:
          As I've mentioned earlier I have implemented the pseudocode from page 734 in the book.
          The problem is that my net is unable to learn the given examples as can be seen from the simple visualization I print out..

          Cheers,
          Matan.
        Your message has been successfully submitted and would be delivered to recipients shortly.