As could cross validation in a neural network NarX?

I would like to know how to raise the cross-validation for NarX neural network, in this way, not destroy correlations. I have a set of 1136 data and 1136 data input targets.

thank you very much

Answer 1

[Shashank Prasanna]

You may want to use DIVIDEBLOCK instead of the default dividerand. DIVIDEBLOCK will maintain correlation since it doesn't shuffle the data but takes block of it for crossvalidation:

http://www.mathworks.com/help/nnet/ug/improving-generalization.html

http://www.mathworks.com/help/nnet/ref/divideblock.html