Questions about time series prediction using nstool

Hi, i'm new to matlab and Neural network. I'hv been fiddling with the nstool- time series prediction for awhile.

I'm confused with the number of delays specified in the nstool.

1. Let's say delay is 2, may i know what it actually means?

2. How do i get the predicted values afterwards?

Answer 1

[Greg Heath]

Cannot help with nstool. However I have posted tens of command line examples in the NEWSGROUP and ANSWERS. Just add greg to your list of search words, e.g.

 greg narxnet

inputdelay = id >= 0 (nonnegative) means that

a. The current output depends on the input id timesteps earlier
b. The current input helps predict the output id timesteps into the future

feedbackdelay = fd > 0 (positive) means that

a. The current output depends on the past output fd timesteps ago
b. The current output helps predict the output fd timesteps into the future

The best approach is to use the command line searches

 help narxnet
 doc narnet

followed by a mathworks search using "greg". For example

 greg timedelaynet

or

 greg narnet

Hope this helps.

Thank you for formally accepting my answer

Greg

[Xiang Jie]

Thanks for your reply! I would like to clarify my understanding for inputdelay.

Part A )

[1 2 3 4 5] - inputs

[6] is the current output that depends on the input 5 timesteps earlier.

Am i right on this?

[Greg Heath]

If you mean ID=5, then y(6) depends only on x(1).