Menu

How to evaluate FER for nnet2 training epoch

Help
JTDamaja
2015-05-27
2015-05-29
  • JTDamaja

    JTDamaja - 2015-05-27

    Hello everyone,
    I've been experimenting lately with nnet and nnet2 trainig.
    In the end of nnet logs for each training epoch I can see FRAME_ACCURACY.
    How to evaluate same FRAME_ACCURACY for epoch of nnet2 training?

     
    • Daniel Povey

      Daniel Povey - 2015-05-27

      I believe it's printed out the the logs in compute_prob_train.XX.log
      and compute_prob_valid.XX.log
      Dan

      On Wed, May 27, 2015 at 7:25 AM, JTDamaja jtdamaja@users.sf.net wrote:

      Hello everyone,
      I've been experimenting lately with nnet and nnet2 trainig.
      In the end of nnet logs for each training epoch I can see FRAME_ACCURACY.
      How to evaluate same FRAME_ACCURACY for epoch of nnet2 training?


      How to evaluate FER for nnet2 training epoch


      Sent from sourceforge.net because you indicated interest in
      https://sourceforge.net/p/kaldi/discussion/1355348/

      To unsubscribe from further messages, please visit
      https://sourceforge.net/auth/subscriptions/

       
      • JTDamaja

        JTDamaja - 2015-05-28

        Thanks for your quick response

        That is, "accuracy is 0.189" == FRAME_ACCURACY?

        LOG (nnet-compute-prob:main():nnet-compute-prob.cc:91) Saw 4000 examples, average probability is -6.45728 and accuracy is 0.189 with total weight 4000
        -6.45728

        Is that so?

         
        • Daniel Povey

          Daniel Povey - 2015-05-28

          yes.

          On Thu, May 28, 2015 at 4:39 AM, JTDamaja jtdamaja@users.sf.net wrote:

          Thanks for your quick response

          That is, "accuracy is 0.189" == FRAME_ACCURACY?

          LOG (nnet-compute-prob:main():nnet-compute-prob.cc:91) Saw 4000 examples,
          average probability is -6.45728 and accuracy is 0.189 with total weight 4000
          -6.45728

          Is that so?


          How to evaluate FER for nnet2 training epoch


          Sent from sourceforge.net because you indicated interest in
          https://sourceforge.net/p/kaldi/discussion/1355348/

          To unsubscribe from further messages, please visit
          https://sourceforge.net/auth/subscriptions/

           
          • JTDamaja

            JTDamaja - 2015-05-29

            Thank you very much

             
MongoDB Logo MongoDB