Jack L.
Guest
|
 Posted:
Sun Dec 26, 2004 5:30 am Post subject:
Re: Weird behaviour in the nLMS code. |
|
|
Jack L. wrote:
| Quote: | Shawn Steenhagen wrote:
"Jack L." <jack_nospam@nospam.dk> wrote in message
news:%tWwd.76207$Vf.3615633@news000.worldonline.dk...
But I get confused because in
Simon Haykin's book "Adaptive Filter Theory", 4th edition, he says
that mu is normalised with respect to the Euclidean norm. If I'm not
wrong, the Euclidean norm is given by: sqrt(sum(w_0[i]*w_0[i])) <--
half MatLab, half C-notation. I don't know what to believe, the book
or what I observe (+ what you say)?
|
Okay, after re-reading the same paragraph in that book many times looking
for something to verify my words I think I get it. Precisely, he writes "the
product vector u(n)e*(n) is normalized with respect to the SQUARED Euclidean
norm of the tap-input vector u(n)". Since the Euclidean norm itself is given
as written above, i.e. ||u(n)|| = sqrt(sum(w_0[i]*w_0[i])), squaring the
whole thing cancels the sqrt, thus we actually divide my*u(n)*e(n) by sum of
w_0[i]*w_0[i]. Am I on the right track now? o_O Is it /that/ single word
'squared' I missed out before?
--
Mvh / Best regards,
Jack, Copenhagen
The email address is for real :) |
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in