[Eeglablist] indipendent component analisys

[Marius Klug]

Hi Dorian, hi Scott,

first of all, thanks for the question and the answer! I was thinking about
the same as well, and the answer I got by now basically was "it's the
math". Your answer is quite interesting in the perspective that in theory
it could be different, right? So, if I understood this right, one could for
example have as many ICs as one has supposed sources in the beamformer for
source reconstruction plus some amount for artifacts and noise? But this
would actually be less useful than having fewer ICs which "sum up" the
small sources to bigger cortical patches, which are actually what we want
to investigate (plus it would probably be a mathematical disaster and not
applicable in acceptable runtime...)? If so this is quite interesting,
because by now I thought, the closer one looks, the better. But at least
concerning EEG, looking closer diffuses the actual signal in a way that
it's not anymore of useful information to us... or at least of different
information, which we're not looking for.

Cheers,
Marius

2015-11-18 20:00 GMT+01:00 Scott Makeig <smakeig at ucsd.edu>:

> Dorian -
>
> Finding the same number of sources as channels makes ICA into a linear
> change of basis problem -- this makes the math simpler and reduces the
> number of assumptions involved in applying the analysis.
>
> We know, of course, that strictly speaking, small-scale potential
> variations in cortex alone are vastly more variegated than the number of
> scalp channels -- but most of these variations will be cancelled out
> through common volume conductance and summation at the scalp electrodes
> (i.e., through destructive phase interference, positive-going and
> negative-going potentials at any time point tending to cancel each other in
> their summation at each scalp electrode channel).
>
> EEG signals are thus dominated by (i.e,. chiefly sum) larger signals
> arising from locally synchronous 'patches' of cortical activity whose
> signals, summed across the disparate patch source activity, thus act as the
> effective (brain) sources of scalp EEG signals. In practice, ICA is of
> interest for brain EEG data analysis because it separates out signals from
> these patches (Delorme et al., 2012 PLoS ONE; Akalin Acar et al,
> Neuroimage, 2015).
>
> Scott Makeig
>
> On Tue, Nov 10, 2015 at 2:11 AM, Dorian Grelli <dorian.grelli at gmail.com>
> wrote:
>
>> Hi guys,
>> I've another questions about ICA. This is more theretical. I am wondering
>> why, after running ICA, we get as many indipendent components as we have
>> channells. I studied a bit of ICA theory in the tutorials ("for dummies"
>> and "not for dummies" that I found on the Internet) but, unfortunately, my
>> background is quite far from math and matrixes and it's difficult for me to
>> digest every detail. I think I get a bit of the theory but the point above
>> is still unclear. Could you help me?
>>
>> Cheers,
>>
>> Dorian
>>
>> _______________________________________________
>> 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
>
> _______________________________________________
> 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
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20151119/ce88b51d/attachment.html>