backpropagation neural network using patternnet take non-Binary training set

I am using Matlab R2012 and wish to design BPNN. Using nprtool, it is possible to design a two layer backpropagation neural network. I have following query:

Can backpropagation neural network using patternnet take the non-Binary dataset as a training set? please guide me. Thank you in advance.

Answer 1

[Greg Heath]

For a c-class classifier, use patternnet with binary targets consisting of columns of the c-dimensional unit matrix eye(c).

Inputs can be real or binary.

There are many examples available. Search

 help patternnet
 doc patternnet

 greg patternnet

Hope this helps.

Thank you for formally accepting my answer

Greg

[Chetana]

Thank you very much. I am working on classification problem. I have train the net and performance of it is attached here for reference.

1. While training, entire dataset is consists of (10 users and 8 samples per user)total 80 images to classify. I have divided entire dataset in two parts- 50 images for training (10 users x 5 samples per user) and 30 images as unseen images (10 users x 3 samples per user).
2. Training dataset is further divided as training, validation, and testing data using 'net.divideFcn = 'dividerand';' and net.divideMode = 'sample'; % Divide up every sample net.divideParam.trainRatio = 80/100; net.divideParam.valRatio = 10/100; net.divideParam.testRatio = 10/100;

Now my query is

1. How to improve classification accuracy for unseen data. 2. Once trained, how to obtained same net performance again ( I know save net command, is it sufficient). pl provide link for any good tutorial. Thank you for your kind cooperation and valuable suggestions.

[Greg Heath]

There are 3 main ways to improve generalization. Although they can be used simultaneously, usually one is sufficient.

 1. Given a training goal, e.g., 

 NMSE = mse(error)/mean(var(target',1)) <= 0.01       

 Minimize the number of hidden nodes

 2. Use a validation stopping subset

 3. Use regularization via
    a. MSEREG 
  or
    b. TRAINBR

Hope this helps,

Greg