Coefficients of the wavelet network

Please,

Can someone tell me what algorithms the Matlab system identification toolbox use to determine the values of the coefficients of the wavelet network? in case, scaling, translation and dilation coefficients in the expansion (as, aw, bs, bw, cs,cw...).

Can anyone tell me the steps of the training algorithm, and if possible point me an article.

Thanks

Answer 1

[Rajiv Singh]

If you have not received this from Dr. Zhang already, please download:

https://skydrive.live.com/redir.aspx?cid=1f5749af4ea080cd&resid=1F5749AF4EA080CD!124

[Emanuel]

Very good!

Thank you very much Rajiv!

Answer 2

[the cyclist]

Have you looked at the documentation here?

http://www.mathworks.com/products/sysid/description4.html

You can see a detailed list of all the functions there.

[Emanuel]

Yes, I've looked. Unfortunately the documentation provided by Matlab for sys. identification toolbox is superficial. I need to know the steps of the training algorithm to wavenet structure used in Matlab.

[the cyclist]

OK. Just to be sure, the list of every function in the toolbox, with at least some detail of what it does, is here: http://www.mathworks.com/help/toolbox/ident/ref/f3-8911.html. (In case you didn't drill down that far.)

If you own the toolbox already, you can edit the functions to see the exact math, or call support to get some help. If you don't own it, and need to know this info to decide on a purchase, then I suggest you contact sales and tell them that.

[Emanuel]

I tried, however, this is an arduous task for a beginner in Matlab. The algorithm programmed by Dr. Zhang involves numerous calls to subroutines with complex functions to understand for a layman in Matlab.

I believe it would be wiser to read the paper from which the learning algorithm was taken.

Answer 3

[Rajiv Singh]

Hi Emanuel, Not that your request is unreasonable, but it is an arduous task to extract out a complete description of the entire estimation algorithm. This is like asking how lsqnonlin works - how much explanation of academic background, fudge factors, numerical algorithms would you include? Where would one draw the line?

I will try looking for some info that would be convenient to share.

[Emanuel]

Somebody help me, please!