[Eeglablist] Fw: question on SICA approach

[Scott Makeig]

Frederica -
  You've misinterpreted -- ICA learns a (channels,channels) unmixing matrix,
so the number of data frames (time points) needed to separate as many
components increases as the square of the number of channels. The faq is a
bit out of date -- in our work with 72-256 channel data in the lab, we have
found that as the channel density becomes high, good ICA solutions typically
require a considerable multiple of the channel number squared (up to 30 or
more for 256-channel data).

For 148-channel data I would like to collect 30*128^2 ~ 650k (~40 min of
data at 256 Hz) ... though it could well be that smaller data sets could
also give useful solutions. I will update the faq to better reflect this,
and will try to do a numerical study to get a more detailed heuristic
understanding.

Scott Makeig

On Feb 16, 2008 2:36 AM, Federica Di Grazia <federicadigrazia at hotmail.com>
wrote:

>
> ----- Original Message ----- *From:* Federica Di Grazia<federicadigrazia at hotmail.com>
> *To:* eeglablist at sccn.ucsd.edu
> *Sent:* Tuesday, January 29, 2008 10:13 PM
> *Subject:* question on SICA approach
>
> I saw the faq on Independent Component Analysis but I couldn't understand
> how many samples(in time) I need to analyze 148 channels with SICA approach?
>
> I saw that it's necessary a number of samples(in time) equal to the square
> of the channels, but for SICA  the sample are represented by the 148
> channels, so I need the square root of 148 as samples(in time)?
>
>
>
> Best Regards,
>
> ____________________________________________________________________
>
> Federica Di Grazia
> Ph. D. Student in Electronics, Automation and Complex Systems Engineering
> Dipartimento di Ingegneria Elettrica Elettronica e dei Sistemi
> Università degli Studi di Catania
> v.le A.Doria 6 - 95125 Catania, Italy
> Tel. +39-095-7382342
> Fax +39-095-330793
> e-mail: fdigra at diees.ing.unict.it
>            federicadigrazia at hotmail.com
> ____________________________________________________________________
>
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to
> eeglablist-request at sccn.ucsd.edu
>



-- 
Scott Makeig, Research Scientist and Director, Swartz Center for
Computational Neuroscience, Institute for Neural Computation, University of
California San Diego, La Jolla CA 92093-0961, http://sccn.ucsd.edu/~scott
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20080217/c97b169a/attachment.html>