Downsampling fsaverage source model for efficient computation

Hi all,

I’m wondering if there is a systematic way of downsampling the source model for fsaverage. I’m using it in source-reconstruction of EEG data with LCMV and it is using up a whole lot of memory (64Gb - on a pretty good machine).

Is it possible to downsample, or is there a file that has less sources than the fsaverage-ico-4-src.fif that I could implement?

Any help would be much appreciated.


This question already has the relevant info: Laptop crashes while trying to compute/plot inverse solution. I suspect my forward model is too intricate. How to address?
…but you say you’re using ico-4 and it’s crashing? I suspect it’s not the forward model that is the problem. Check your script to see if you’re keeping around unneeded variables. For example, don’t load raw data for all subjects all at once; load one subj at a time, process them all the way through, then load the next subj.

If you can’t prevent the crash that way, please post the script and the full traceback and the output of mne.sys_info()