[FieldTrip] problem with ICA

[Hwee Ling Lee]

Hi,

Thanks for suggestion. Actually, prior to running ICA, I did a notch filter
of 50 Hz and also to remove cardioballistic effects based on the ECG
channel.

Does it consider to be mixing the channels?

Cheers,
Hweeling



On 16 January 2014 13:54, Eelke Spaak <eelke.spaak at donders.ru.nl> wrote:

> Hi Hweeling,
>
> To add to Aaron's explanation, you can instruct the algorithm to use a
> subspace projection of your data by specifying cfg.runica.pca = N,
> where N is the rank of your data (in your case 119, it seems).
>
> Best,
> Eelke
>
> On 16 January 2014 13:42, Aaron Schurger <aaron.schurger at gmail.com> wrote:
> > Sounds like you may have done something to your data, like
> > interpolating channels, before you ran ICA. It is OK to filter your
> > data before running ICA, and some other operations are OK too, but if
> > you do anything that mixes activity from different channels in any
> > way, then you can run into problems with ICA (and results from ICA can
> > be invalid).
> > Aaron
> >
> > On Thu, Jan 16, 2014 at 1:22 PM, Hwee Ling Lee <hweeling.lee at gmail.com>
> wrote:
> >> Dear all,
> >>
> >> I've collected data using a 128 channel EEG cap, and I tried to perform
> ICA
> >> on the data. However, I got an error message with fieldtrip on Matlab.
> >> Here's the error message:
> >>
> >> the input is raw data with 127 channels and 1 trials
> >> selecting 123 channels
> >> baseline correcting data
> >> scaling data with 1 over 148.247820
> >> concatenating data.
> >> concatenated data matrix size 123x2789000
> >> starting decomposition using runica
> >>
> >> Input data size [123,2789000] = 123 channels, 2789000 frames/nFinding
> 123
> >> ICA components using logistic ICA.
> >> Decomposing 184 frames per ICA weight ((15129)^2 = 2789000 weights,
> Initial
> >> learning rate will be 0.001, block size 75.
> >> 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-07 or after 512 steps.
> >> Online bias adjustment will be used.
> >> Removing mean of each channel ...
> >> Final training data range: -3.46556 to 6.39436
> >> Computing the sphering matrix...
> >> Starting weights are the identity matrix ...
> >> Sphering the data ...
> >> Beginning ICA training ...
> >> Data has rank 119. Cannot compute 123 components.
> >> the call to "ft_componentanalysis" took 148 seconds
> >>
> >> Could someone please let me know what went wrong?
> >>
> >> Thanks!
> >>
> >> Cheers,
> >> Hweeling
> >>
> >>
> >> _______________________________________________
> >> fieldtrip mailing list
> >> fieldtrip at donders.ru.nl
> >> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
> >
> >
> >
> > --
> > Aaron Schurger, PhD
> > Senior researcher
> > Laboratory of Cognitive Neuroscience
> > Brain-Mind Institute, Department of Life Sciences
> > École Polytechnique Fédérale de Lausanne
> > Station 19, AI 2101
> > 1015 Lausanne, Switzerland
> > +41 21 693 1771
> > aaron.schurger at epfl.ch
> > http://lnco.epfl.ch/
> >
> > _______________________________________________
> > fieldtrip mailing list
> > fieldtrip at donders.ru.nl
> > http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>



-- 
=================================================
Dr. rer. nat. Lee, Hwee Ling
Postdoc
German Center for Neurodegenerative Diseases (DZNE) Bonn

Email 1: hwee-ling.lee<at>dzne.de
Email 2: hweeling.lee<at>gmail.com

https://sites.google.com/site/hweelinglee/home

Correspondence Address:
Ernst-Robert-Curtius Strasse 12, 53117, Bonn, Germany
=================================================
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20140116/c5dfd309/attachment-0002.html>