NAR network vs NARX network? Error weights in preparets?

I need to know the difference between the two different network which are NAR network and NARX network. Besides the point NAR is based on past values and NARX is based on past values with external input.

I also need to know if there is a difference when I add in the error weight in the preparets? Will it produce better prediction and what will it do?

Lastly for a 48 timesteps, how do I determine the ideal hiddenLayers?

Thanks in advance for helping.

Answer 1

[Greg Heath]

Narx can also include present values of the input.

If the cross=correlations between input and output are significant, Narx should be better.

Choose H by trial and error. A trial run with the default H=10 is not a bad idea.

Search for some of my relevant code

 greg narxnet Hub

Hope this helps.

Thank you for formally accepting my answer

Greg