Memory Error

✉️ Mailing List Archive (read-only)
1

External Email - Use Caution

Dear MNE users,

I am trying to do something similar that has been done in this example:
https://mne.tools/stable/auto_examples/inverse/plot_psf_ctf_vertices.html#sphx-glr-auto-examples-inverse-plot-psf-ctf-vertices-py

Except, I would like to compute my forward and Inverse operator with free
orientation instead of being fixed or perpendicular to the cortical
surface. However, I am getting this following error once I try to compute
it. Is there any workaround to skip this memory error??

Best,
Dip

*Error: *

*Traceback (most recent call last): File
"/media/diptyajit/Elements/Pycharm_Aud_Cog/plot_psf_ctf_vertices.py", line
53, in <module> snr = 3.0 File "<decorator-gen-313>", line 21, in
make_inverse_resolution_matrix File
"/home/diptyajit/anaconda3/lib/python3.7/site-packages/mne/minimum_norm/resolution_matrix.py",
line 60, in make_inverse_resolution_matrix resmat =
invmat.dot(leadfield)MemoryError: Unable to allocate array with shape
(56310, 56310) and data type float64*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20200625/a0bfba59/attachment.html

2

External Email - Use Caution

hi,

try reducing the size of your source space or find a machine with more RAM.
It seems you need at least 25GB here.

HTH
Alex

3

External Email - Use Caution

Hi alex,

Thanks for the suggestions. I tried with reduced source space('oct6'). I
can compute the resolution matrix but it doesn't match the size of the no.
of vertices of the original source space. Therefore, I am getting an error
while saving it as source estimate object .
Details:
original source space: 7503 locations (3 dipoles per each location)
Dimension of Inverse Matrix: (22509, 187)
Dimensions of resolution matrix: 22509 by 22509.
ValueError: Number of vertices (7503) and stc.shape[0] (22509) must match

However, I am expecting a resolution matrix with size of 7503*7503. Is the
resolution matrix being calculated each dipole location instead of taking
of normal of 3 components?? Any idea?

best,
Dip

4

External Email - Use Caution

hi,

cc Olaf

it could be a bug. Can you open an issue so we can have a look at it?

Alex

5

External Email - Use Caution

Hi,

when you compute the resolution matrix R for n locations and free orientations, then R will have dimension (3*n)x(3*n), since every dipole can affect every other dipole. If you want to put this into an STC, you will have to combine source orientations (e.g. RMS) in a separate step.
If this doesn't answer your question, please send me some more details about what exactly you are doing, and I'll look into it.

Best wishes,

Olaf