ICA - problems

Discussions
1

Hi,

I have problems viewing the ICA components after running the script into MatLab. This is the message Brainstorm gives me (see attachment) and afterwhat I don’t have any other choice than completely shut down the “terminal” because both Matlab and Brainstorm are “freezed”.

I would really appreciate if we could explain me what I am doing wrong and how it can’t fix the problem.

Thank you very much,

Véronique M.

2

Hi Veronique,

There are two problems here:

  1. The ICA algorithm produces something that is not meaningful and apparently contains complex values. I don’t have much control over this, as it is directly calling the EEGLAB functions runica.m.
    What I could add, it I knew what was wrong in your case, would be to add a more meaningful error message instead of this abrupt crash, but I don’t think I can fix the issue.
    Try with different parameters, starting from setting the number of ICA components to something lower than the total number of sensors.
    Do you get the same problem with both Infomax and Jade ?
    While Infomax is running, you should be able to track the optimization in the Matlab command window (the values lrate and wchange should go down at almost every iteration). If they stay around the same value, there is something wrong with the algorithm or the data in input. Maybe you could try to post a message on the EEGLAB mailing list…

  2. Matlab freezes when trying to call something graphical. This can be a problem related with your graphics card driver and OpenGL support. See the options to test in the following thread:
    http://neuroimage.usc.edu/forums/showthread.php?2350

Cheers,
Francois

3

Hi!

Thanks for the fast response!

Strangely after closing MatLab and Brainstorm, everything went well (which is weird because the optimization track in Matlab was exactly the same thing as when I got the crash except I went a lot faster when it finally work).
I did it with Infomax.

Cheers,
Véronique M