[FieldTrip] SVD and ICA

Peter Goodin pgoodin at swin.edu.au
Sat Jan 11 13:47:56 CET 2014


Hi Federico,

Seeing as how you've stated using SSS, I'm going to assume you're using a neuromag system which has decreased the dimensionality of your data already through the extraction of "B-out" components.  There are two options - the first is to use runica and and reduce the number of components to ~70 through PCA (covered in a couple of posts on this list). The second is to use the fastica algorithm which will automagically calculate the optimal amount of components to be extracted from the data. 

ICA will typically give far more components than SSP will projectors as ICA a model free method (so includes things such as EOG + ECG + EMG + external arefact + brain components). SSP however is model based and will only return projectors based on input (such as examples of eye blinks / ECG). 

Hope this helps, 

Peter

__________________________
Peter Goodin,
BSc (Hons), Ph.D Candidate.

Brain and Psychological Sciences Research Centre (BPsych)
Swinburne University,
Hawthorn, Vic, 3122

Monash Alfred Psychiatry Research Centre (MAPrc)
Level 4, 607 St Kilda Road,
Melbourne 3004

________________________________________
From: fieldtrip-bounces at science.ru.nl [fieldtrip-bounces at science.ru.nl] on behalf of Federico Grande [fgrande at cbs.mpg.de]
Sent: Saturday, 11 January 2014 10:37 PM
To: Diego Lozano; FieldTrip discussion list
Subject: Re: [FieldTrip] SVD and ICA

Hi Diego,

I'm sorry, I was probably not clear enough. When you uses SSP (Signal Space Projection), to process the rawdata, it projects the data in 8 or 10 PCA, but when you uses SSS in the rawdata, it has a much higher amount of components , around 150 or almost 200. That is the reason that makes me having no idea about how should I reduce it.

Cheers,

Federico

----- Mensaje original -----
De: "Lozano Soldevilla, D. (Diego)" <d.lozanosoldevilla at fcdonders.ru.nl>
Para: "FieldTrip discussion list" <fieldtrip at science.ru.nl>
Enviados: Viernes, 10 de Enero 2014 18:27:17
Asunto: Re: [FieldTrip] SVD and ICA

Hi Federico,
I don't follow you. What's SSP? In any case, what I explained it's the way that I know to reduce data dimensionality prior ICA computation. I don't know a procedure to know optimal number of PCA component but here 25 were used: http://www.ncbi.nlm.nih.gov/pubmed/19699307

best,

Diego

----- Original Message -----
> From: "Federico Grande" <fgrande at cbs.mpg.de>
> To: "Diego Lozano" <d.lozanosoldevilla at fcdonders.ru.nl>, "FieldTrip discussion list" <fieldtrip at science.ru.nl>
> Sent: Friday, 10 January, 2014 6:14:28 PM
> Subject: Re: [FieldTrip] SVD and ICA
> Aham, that is what I've done, do the runica method, but I didn´t use
> the parameter pca: pca are not principal component analysis associated
> to SSP? I have used SSS (signal space separation) instead of SSP. It
> would work also? And also I don´t know what number of components do I
> want to reduce my data. How can I know which is the optimal number?
>
> Thank you Diego,
>
> Federico
>
> ----- Original Message -----
> From: "Lozano Soldevilla, D. (Diego)"
> <d.lozanosoldevilla at fcdonders.ru.nl>
> To: "FieldTrip discussion list" <fieldtrip at science.ru.nl>
> Sent: Friday, January 10, 2014 4:13:13 PM
> Subject: Re: [FieldTrip] SVD and ICA
>
> Hi Federico,
>
> You might want to have a look to the different ICA algorithms
> ft_componentanalysis has and see how to choose the proper option. For
> example, if you select cfg.method='runica' then cfg.runica.pca =
> number of components you want to reduce your data.
>
> Check help ft_componentanalysis for details
>
> best,
> Diego
>
>
> ----- Original Message -----
> > From: "Federico Grande" <fgrande at cbs.mpg.de>
> > To: fieldtrip at science.ru.nl
> > Sent: Friday, 10 January, 2014 3:54:56 PM
> > Subject: [FieldTrip] SVD and ICA
> > Hello everyone,
> >
> > In order to remove the artefacts like blink eyes or hearbeat, I
> > wanted
> > to apply ICA to my data. I've been told that is better to apply
> > first
> > SVD and then ICA, but I don't really know how to apply it. What do
> > you
> > recommend me in order to do it? I've not found any tutorial for
> > doing
> > it. All help and information ins greatly welcomed.
> >
> > Thank you very much,
> >
> > King Regards,
> >
> > Federico Grande
> > _______________________________________________
> > fieldtrip mailing list
> > fieldtrip at donders.ru.nl
> > http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>
> --
> PhD Student
> Neuronal Oscillations Group
> Donders Institute for Brain, Cognition and Behaviour
> Centre for Cognitive Neuroimaging
> Radboud University Nijmegen
> NL-6525 EN Nijmegen
> The Netherlands
> http://www.ru.nl/people/donders/lozano-soldevilla-d/
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip

--
PhD Student
Neuronal Oscillations Group
Donders Institute for Brain, Cognition and Behaviour
Centre for Cognitive Neuroimaging
Radboud University Nijmegen
NL-6525 EN Nijmegen
The Netherlands
http://www.ru.nl/people/donders/lozano-soldevilla-d/

_______________________________________________
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




More information about the fieldtrip mailing list