Spatio-temporal clustering

✉️ Mailing List Archive (read-only)
1

Dear mne-python experts,
We are trying to run spatio-temporal cluster stats for my experiment. I
tried to follow the 'plot_cluster_stats_spatio_temporal.py' example with 5
participants (and two conditions). I am trying to run my analyses in my
iMac Os 10.6.8 with 4 Gb of RAM. We are using the last update of mne (0.5)
and we constantly find the following error.

...
[done]
    Reading a surface...
[done]
    8 BEM surfaces read
-- number of connected vertices : 20484
Clustering.
Python(17084) malloc: *** mmap(size=557166592) failed (error code=12)
*** error: can't allocate region
*** set a breakpoint in malloc_error_break to debug
Bus error

any suggestion?

(By the way, with 2 participants the script works...)

N
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20130226/732b5925/attachment.html
-------------- next part --------------
04
05
06
08
09
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Spatio_temp.py
Type: application/octet-stream
Size: 4657 bytes
Desc: not available
Url : http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20130226/732b5925/attachment.obj

2

Hi Nicola,

It seems like you don't have sufficient memory to run the analysis. It
tries to allocate a memmap with 557166592 elements, which corresponds to
4GB (8 bytes per element).

Did you use parallel processing? If so, try running it with n_jobs=1,
this should reduce the memory consumption At the moment you need almost
n_jobs times the memory when using parallel processing. Hopefully this
will be improved in the future.

Best,

Martin

3

hi Nicola,

what you can try is filter and downsample your time series to have less data
or try casting the stc.data to float 32 to have single precision.

HTH
Alex