[Eeglablist] running ICA

[Makoto Miyakoshi]

Dear Karlo,

I saw it when my data were very dirty or rank-reduced.

Run following codes and tell me what you see.
1. figure; bar(std(EEG.data(:,:),0,2));
Use data reader tool (find it from matlab figure icons). If you see any bar
standing out, read its channel number (let's say it is channel N) and run
next one.
2. figure; plot(EEG.data(N,:))
Also run
3. figure; bar(mean(EEG.data(:,:)));

Makoto

2013/5/6 karlo gonzales <thats_karlo at yahoo.com>

> Dear Friends,
>
> I faced a problem with running ICA, that required your expertise
> and guides. I am simply using:  EEG = pop_runica(EEG,'icatype','runica'); but
> it seems because of the nature of my data (rank , Gaussian, or ....), ICA
> algorithm needs to lower learning rate frequently after few steps (please
> see the below). size of my data is  >> size(EEG.data)       >> ans=  62
> 4820   159.
>
> - how do you check rank of your EEG data? (rank(EEG.data)  gives error for
> 3d matrix )
> - what could be wrong here or what i need to check before running ICA?
> - May i ask you  that how do you choose your ICA options (e.g., 'pca'  ,
> 'extended1' or ...)?
>
> Many thanks,
> Karlo
>
>
>
> Input data size [62,766380] = 62 channels, 766380 frames/nFinding 62 ICA
> components using logistic ICA.
> Decomposing 199 frames per ICA weight ((3844)^2 = 766380 weights, Initial
> learning rate will be 0.001, block size 68.
> Learning rate will be multiplied by 0.9 whenever angledelta >= 60 deg.
> More than 32 channels: default stopping weight change 1E-7
> Training will end when wchange < 1e-007 or after 512 steps.
> Online bias adjustment will be used.
> Removing mean of each channel ...
> Final training data range: -456.566 to 562.934
> Computing the sphering matrix...
> Starting weights are the identity matrix ...
> Sphering the data ...
> Beginning ICA training ...
> Lowering learning rate to 0.0009 and starting again.
> step 1 - lrate 0.000900, wchange 141.51904569, angledelta  0.0 deg
> Lowering learning rate to 0.00081 and starting again.
> step 1 - lrate 0.000810, wchange 159.49273116, angledelta  0.0 deg
> Lowering learning rate to 0.000729 and starting again.
> step 1 - lrate 0.000729, wchange 143.22748820, angledelta  0.0 deg
> Lowering learning rate to 0.0006561 and starting again.
> step 1 - lrate 0.000656, wchange 136.88732118, angledelta  0.0 deg
>
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to
> 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
>



-- 
Makoto Miyakoshi
Swartz Center for Computational Neuroscience
Institute for Neural Computation, University of California San Diego
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20130509/f014125e/attachment.html>