Question on Narxnet with open-loop

I have a question about Narxnet.

In the narxent, we see that: In closed-loop, we only use inputs and training the weights to inputs and to the feedback output. And in removedelay(), we use current input and output to predict the next step output,

Then may I know what is the purpose of using current(/past) input and current output to predict current output? Can someone explain the diagram here? I am kind of confused.

Thanks

[Jayaram Theegala]

The output 'y' at every time step is determined by the current input as well as the output. In other words, the purpose of the past inputs and the output is to represent the nonlinear autoregressive model and thereby predicting the output by taking past outputs into account. I think you may find the following information about, NARX feedback neural networks useful:

https://www.mathworks.com/help/nnet/ug/design-time-series-narx-feedback-neural-networks.html

Answer 1

[Greg Heath]

WARNING:

Cannot use a closeloop 0 lag delay for feedback.

Although you can use an openloop 0 feedback delay, trying to close the loop will not yield a valid result.

I don't have MATLAB installed so I cannot demonstrate. In particular, I don't remember if you will get an error message or not.

You can try to search in the NEWSGROUP and ANSWERS, however my advice is to just try a trivial example.

Hope this helps.

Thank you for formally accepting my answer

Greg