[Eeglablist] Memory problem running ICA

[Petr Janata]

Hi Greg,

Your problem may be that current versions of Matlab only have 32-bit 
addressing of memory space, and actually address only 31 bits (2 Gb) of 
memory.  So, 6 Gb doesn't do you any good in the context of a single 
Matlab process.

Here are a couple of relevant Mathworks pages:

http://www.mathworks.com/support/solutions/data/27931.shtml
http://www.mathworks.com/support/solutions/data/35322.shtml

Best,
Petr


On Mar 24, 2004, at 2:57 PM, Greg Appelbaum wrote:

> Hi there,
> I'm wondering if anyone has a suggestion for a memory error I am 
> running
> into when computing ICA weights using runica.  I am using a Linux 
> system
> with 6GB of RAM to analyze a 500Mb continuous 148 channel session of 
> MEG
> data I have.  I have checked to see that the memory resources are 
> intact
> (plenty of RAM left when viewing from top and no use of swap space) 
> and have
> tried unsetting the precompute ICA option in eeglab.  Nonetheless I
> continually run into the error message listed below.  Any suggestions 
> would
> be greatly appreciated.
>
> Your
>
> Greg Appelbaum
>
>
>
> Input data size [148,402300] = 148 channels, 402300 frames.
> Finding 148 ICA components using logistic ICA.
> Initial learning rate will be 0.001, block size 64.
> Learning rate will be multiplied by 0.9 whenever angledelta >= 60 deg.
> More than 32 channels: default stopping weight change 1E-7
> Training will end when wchange < 1e-07 or after 512 steps.
> Online bias adjustment will be used.
> Removing mean of each channel ...
> Final training data range: -6.90942e-11 to 6.27449e-11
> Computing the sphering matrix...
> ??? Out of memory. Type HELP MEMORY for your options.
>
> Error in ==> /local/software/matlab-6.5/toolbox/matlab/elmat/repmat.m
> On line 54  ==>   B = A(mind,nind);
>
> Error in ==> /local/software/matlab-6.5/toolbox/matlab/datafun/cov.m
> On line 61  ==>   xc = x - repmat(sum(x)/m,m,1);  % Remove mean
>
> Error in ==> /es/l/la/lappelba/eeglab4.311/functions/runica.m
> On line 718  ==>   sphere = 2.0*inv(sqrtm(cov(data'))); % find the
> "sphering" ma
> trix = spher()
>
> Error in ==> /es/l/la/lappelba/eeglab4.311/functions/pop_runica.m
> On line 257  ==>                 [EEG.icaweights,EEG.icasphere] = 
> runica(
> tmpdat
> a, 'lrate', 0.001 );
>
> Error in ==> /es/l/la/lappelba/matlab_code/SOBI_code/ICAroutine.m
> On line 31  ==>     EEG = pop_runica(EEG, 'runica');
>
> _______________________________________________
> Eeglablist mailing list Eeglablist at sccn.ucsd.edu
> Eeglablist page: http://www.sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to eeglablist-unsub at sccn.ucsd.edu
> To switch to digest mode, send an empty email to 
> eeglablist-digest at sccn.ucsd.edu
>
>
------------------
Petr Janata, Ph.D.
Research Assistant Professor
Department of Psychological and Brain Sciences, and
Center for Cognitive Neuroscience
Dartmouth College
Hanover, NH 03755

phone: 603 646 0062
fax: 603 646 1419
e-mail: petr.janata at dartmouth.edu
WWW: http://atonal.dartmouth.edu/~petr

EFFECTIVE JULY 1, 2004
My affiliation will be UC Davis.
pjanata at ucdavis.edu