[Eeglablist] EEG, ALLEEG, and OutOfMemory problem with ICA

Scott Makeig smakeig at gmail.com
Thu Mar 17 13:37:24 PST 2005


The problem is that runica() computes activations and back-projections 
before it exits, to order the components by variance accounted for and 
to pick  polarities. I am now testing a version of runica() that 
minimizes the space required for these computations and plan to release 
it quite soon. I will work on the binary ICA program as well, though 
this may already have less memory overhead than Matlab, e.g. if run 
separately when no Matlab session is active.

Ultimately, however, the ~2 GB limit of 32-bit processes contstrains 
efficient computations on very large datasets. We are now ordering 
64-bit Opterons with more RAM (8GB) to be able to process very large EEG 
and EEG/MEG datasets.

Scott Makeig



Andre Achim wrote:

>I realised that ICA, ran from the graphics interface, runs well with an EEG
>file of reasonable length, up to nearly the end.  I also realised that it
>gets to about the same point if I delete (clear) ALLEEG, but then it cannot
>perform the latest computation stages.  I could not find, however, in which
>.m file these latest computation stages are defined.  Since EEG and ALLEEG
>seem identical with a single data file, I would like to change the reference
>to ALLEEG for a reference to EEG (or conversely).   Alternately, it should
>be possible to store the intermediate results, exchange EEG for ALLEEG and
>then run the final set of command lines.
>
> 
>
>Is that possible?  How?
>
> 
>
>Thank you
>
> 
>
>André Achim 
>
>
>  
>
>------------------------------------------------------------------------
>
>_______________________________________________
>eeglablist mailing list eeglablist at sccn.ucsd.edu
>http://sccn.ucsd.edu/mailman/listinfo/eeglablist
>Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
>To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu
>





More information about the eeglablist mailing list