[Eeglablist] ICA: slow and then crash on one file
Scott Makeig
smakeig at gmail.com
Sun Jun 5 06:35:43 PDT 2011
... else use Matlab rank() to find the dimension of the data. Often, this
might be, e.g., 63 instead of 64 because of redundancy produced by
referencing. If, for example, referencing was to digitally linked mastoids,
(A1+A2)/2, but both mastoid channels were retained:
A1 - (A1+A2)/2 A2 - (A1+A2)/2
= (A1-A2)/2 = (A2-A1)/2
then these two channels do not contain different information (one is the
negative of the other)... So the rank of the data must be (at most) one less
than the number of channels. Here, remove a redundant channel or use PCA to
reduce the rank by 1 before applying ICA.
Note: The Matlab rank() function returned wrong results in some early
version 7 instances.
Scott Makeig
On Sat, Jun 4, 2011 at 8:30 PM, Arnaud Delorme <arno at ucsd.edu> wrote:
> Dear Jeff,
>
> there might be a blank channel or some linear dependencies between
> channels. Try reducing the rank of your data using PCA to see if ICA
> converges better. For instance, if you have 64 channels, try adding the
> option "'pca', 60" or even "'pca', 30".
> Best regards,
>
> Arno
>
>
> _______________________________________________
> 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 & Adj. Prof. of
Neurosciences, University of California San Diego, La Jolla CA 92093-0559,
http://sccn.ucsd.edu/~scott
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20110605/fd417ff6/attachment.html>
More information about the eeglablist
mailing list