[Eeglablist] ICA and covariance matrix for CSP

Wed May 15 11:04:36 PDT 2013

Hi Christian,
Ok. Thank you very much. 
Also I will try to avoid the 'leave one out' cross-validation scheme as that will be a lot of ICA computation!
Bethel
________________________________________
From: Christian Kothe [christiankothe at gmail.com]
Sent: 15 May 2013 18:41
To: Bethel Osuagwu; eeglablist at sccn.ucsd.edu
Subject: Re: [Eeglablist] ICA and covariance matrix for CSP

Hi Bethel,

no, I mean compute the two covariance matrices from EEG.icaact instead of EEG.data (that's the time series in component space instead of channel space). You'd do the ICA on your entire training set (both classes).

Note that if you do a k-fold cross-validation, you need to do the ICA k times (once for each training set), otherwise you're double-dipping<http://www.nature.com/neuro/journal/v12/n5/abs/nn.2303.html>.

Best,
Christian

On Tue, May 14, 2013 at 3:52 PM, Bethel Osuagwu <b.osuagwu.1 at research.gla.ac.uk<mailto:b.osuagwu.1 at research.gla.ac.uk>> wrote:
Dear Christian,
Thank you for you reply.
I not quit sure of how to go about computing CSP from ICA component activation. I am using CSP by Ramoser( http://www.cs.hmc.edu/~keller/eeg/Ramoser.pdf). Two classes of data are required by this method and I planning to use the CSP for classification of BCI data. Do I need to compute ICA separately on each of the data classes to obtain separate EEG.icawinv? Is it  EEG.icawinv that you are referring to as components activation?

Any guide or a paper will be appreciated?

Thanks
Bethel

________________________________________
From: Christian Kothe [christiankothe at gmail.com<mailto:christiankothe at gmail.com>]
Sent: 02 May 2013 05:53
To: Arnaud Delorme
Cc: Bethel Osuagwu; eeglablist at sccn.ucsd.edu<mailto:eeglablist at sccn.ucsd.edu>
Subject: Re: [Eeglablist] ICA and covariance matrix for CSP

Dear Bethel,

note that there is no need to back-project to channels after artifact rejection -- you can apply CSP directly to the component activations.

Christian

PS: For computing a regularized covariance matrix see the function cov_shrink in BCILAB.

On Wed, May 1, 2013 at 9:30 PM, Arnaud Delorme <arno at ucsd.edu<mailto:arno at ucsd.edu><mailto:arno at ucsd.edu<mailto:arno at ucsd.edu>>> wrote:
Dear Bethel,

does the warning also show up when you convert you data matrix to double precision (double function of Matlab).
Best,

Arno

On 25 Apr 2013, at 06:30, Bethel Osuagwu wrote:

Thank you Makoto for your reply.
Reducing the number of channels stops the warning, but I did not check if the number of channels I eliminated equals the number of components removed. I will check if that is true although I think its true.
Also I have been told from FieldTrip that the problem can be solved by regularizing the (inverse)covariance matrix(adding some estimated noise to the diagonals) but am not sure exactly how much noise and the effect of adding this noise.
Thanks
Bethel
________________________________________
From: Makoto Miyakoshi [mmiyakoshi at ucsd.edu<mailto:mmiyakoshi at ucsd.edu><mailto:mmiyakoshi at ucsd.edu<mailto:mmiyakoshi at ucsd.edu>>]
Sent: 24 April 2013 21:14
To: Bethel Osuagwu
Cc:
Subject: Re: [Eeglablist] ICA and covariance matrix for CSP

Dear Bethel,

Sounds like it is because of reduced rank. Please try that you also discard the same number of channels after IC rejection to see what happens.

Makoto

2013/4/23 Bethel Osuagwu <b.osuagwu.1 at research.gla.ac.uk<mailto:b.osuagwu.1 at research.gla.ac.uk><mailto:b.osuagwu.1 at research.gla.ac.uk<mailto:b.osuagwu.1 at research.gla.ac.uk>><mailto:b.osuagwu.1 at research.gla.ac.uk<mailto:b.osuagwu.1 at research.gla.ac.uk>>>
Hi Everyone,

I have noticed that after using ICA to remove artifact in my signal, I can no longer compute the covariance matrices needed for common spatial pattern without matlab giving me warning regarding the accuracy of the estimated matrix. Indeed the estimated matrix is not accurate. This does not happen if I use only a small subset of my channels after removing noise with ICA.  If I do not use ICA at all the warning disappears.

I want to ask if anybody knows why ICA causes this issue and how I can fix it?

Thanks a lot.
Bethel Osuagwu
_______________________________________________
Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu<mailto:eeglablist-unsubscribe at sccn.ucsd.edu><mailto:eeglablist-unsubscribe at sccn.ucsd.edu<mailto:eeglablist-unsubscribe at sccn.ucsd.edu>><mailto:eeglablist-unsubscribe at sccn.ucsd.edu<mailto:eeglablist-unsubscribe at sccn.ucsd.edu>>
For digest mode, send an email with the subject "set digest mime" to eeglablist-request at sccn.ucsd.edu<mailto:eeglablist-request at sccn.ucsd.edu><mailto:eeglablist-request at sccn.ucsd.edu<mailto:eeglablist-request at sccn.ucsd.edu>><mailto:eeglablist-request at sccn.ucsd.edu<mailto:eeglablist-request at sccn.ucsd.edu>>

--
Makoto Miyakoshi
Swartz Center for Computational Neuroscience
Institute for Neural Computation, University of California San Diego
_______________________________________________
Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu<mailto:eeglablist-unsubscribe at sccn.ucsd.edu><mailto:eeglablist-unsubscribe at sccn.ucsd.edu<mailto:eeglablist-unsubscribe at sccn.ucsd.edu>>
For digest mode, send an email with the subject "set digest mime" to eeglablist-request at sccn.ucsd.edu<mailto:eeglablist-request at sccn.ucsd.edu><mailto:eeglablist-request at sccn.ucsd.edu<mailto:eeglablist-request at sccn.ucsd.edu>>

_______________________________________________
Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu<mailto:eeglablist-unsubscribe at sccn.ucsd.edu><mailto:eeglablist-unsubscribe at sccn.ucsd.edu<mailto:eeglablist-unsubscribe at sccn.ucsd.edu>>
For digest mode, send an email with the subject "set digest mime" to eeglablist-request at sccn.ucsd.edu<mailto:eeglablist-request at sccn.ucsd.edu><mailto:eeglablist-request at sccn.ucsd.edu<mailto:eeglablist-request at sccn.ucsd.edu>>