MATLAB Answers

Soumyadip

Deciding number of epochs for pattern recognition

Asked by Soumyadip
on 20 Jun 2013

How to predict the number of epochs a neural network needs for pattern recognition ? Also, how do I decide about the other parameters such as learning rate, momentum constant, etc? Is it by trial and error?

I am working on an OCR project and the images for my training are like the ones shown in the question below :-

http://www.mathworks.in/matlabcentral/answers/78841-size-and-alignment-of-a-digit-in-an-image-for-neural-network-training

  0 Comments

1 Answer

Answer by Greg Heath
on 1 Jul 2013
 Accepted answer

1. Use as many default values as possible. 2. To minimize the number of hidden nodes, H, outer loop over H = 0:dH:Hmax and inner loop over ntrial = 1:Ntrials the number of weight initialization trials for each value of H.

Hope this helps

Thank you for formally accepting my answer

Greg

  0 Comments


Discover what MATLAB® can do for your career.

Opportunities for recent engineering grads.

Apply today