improve the performance of nprtool

I used the neural network toolbox ( nprtool ) for classifying my objects. i used 75% of data for training and 15% for both validation and testing.also i considered 50 neurons for hidden layers. the progress stops because of validation checks (at 6). how can i improve the performance of this network? i couldn't find out how to change validation check or gradient ,....if you have any suggestion i will be very appreciate to hear that.

Answer 1

[Greg Heath]

Insufficient information

 Which of the MATLAB classification example datasets are you using?

 help nndatasets
 doc nndatasets

 Number of classes  c =?
 Input vector dimensionality I = 1
 Number of examples  N = ?
 [ I N ] = size(input)
 [ O N ] = size(target)% O = c
 Default 70/15/15 data division? (75/15/15 doesn't add to 100)

Some problems require multiple(e.g., 10) designs for every value of hidden nodes that are tried.

For example, search the NEWSGROUP and ANSWERS using

 greg patternnet Ntrials

 Sorry I can't give you much advice on how to optimize the use of nprtool. However, consulting my command line code should be more than worthwhile.

Hope this helps.

Thank you for formally accepting my answer

Greg.

[maryam]

thank you for your reply. my input data is a 8192*300 matrix which i'd like to classify it into 3 groups. i used default parameters in nprtool: a two layer feed-forward network and trainscg as trainer. my question is how i can define different conditions for validation check,iterations,...?

[Greg Heath]

I recommend substantially reducing the 300-dimensional inputs using PLS. Then consult some of my classification posts as recommended above.

Hope this helps.

Greg

Answer 2

[smriti garg]

Hello Sir,

As in the GUI of nprtool there is a option of 'retrain' to achieve better performance....Similarly, when using the advanced script of nprtool is there some method to retrain the 'net' function to achieve better peroformance.

Please suggest some solution.

Thanx in advance for help.