Lag Space in NARX NN

I'm trying to model a MIMO sytem (two inputs and two outputs) with a NARX neural net. My problem is that no matter which combination of past inputs and outputs I use, I keep getting a poor performance when simulating it in closed loop.

I've been reading about using the correlation functions to get a good estimation of the needed lags by plotting the results of these functions but I don't quite understand how to use them. I would really appreciate it, if somebody could explanin it to me or post some code to obtain the lags.

Answer 1

[Greg Heath]

Many code examples have been posted. Search on

 greg nncorr narxnet

Hope this helps

Thank you for formally accepting my answer

Greg

[Yevgeniy Arabadzhi]

I searched " greg nncorr" and it brought me here. Wondering when will I find this famous example.

[Greg Heath]

Search the NEWSGROUP 1st.That is where I tend to post tutorial types.

If you want to shorten the list, try adding tutorial to the search words.