[FieldTrip] Two versions of matlab yield different output from runica

[Jörn M. Horschig]

Dear Roey,

 

that is indeed true. ICA (at least most versions like runica) use some random initialization of the weights, which is realized by using a random number generator. You can set the random seed manually in FieldTrip/EEGLAB to verify whether you get the same results for different Matlab versions (as long as the algorithms for the random number generator has not changed). Not sure if results will be same across different operating systems though. 

 

Best,

Jörn

 

PS: I vaguely remembered some bug in EEGLAB (which’s algorithm FieldTrip uses), so I looked it up on Bugzilla: http://bugzilla.fieldtriptoolbox.org/show_bug.cgi?id=2585 so that one seems to be resolved.

 

--

 

Jörn M. Horschig, PhD, Software Engineer

 <http://www.artinis.com/> Artinis Medical Systems  |  +31 481 350 980 

 

From: fieldtrip-bounces at science.ru.nl [mailto:fieldtrip-bounces at science.ru.nl] On Behalf Of Roey Schurr
Sent: Thursday, July 16, 2015 10:15 AM
To: FieldTrip discussion list
Subject: Re: [FieldTrip] Two versions of matlab yield different output from runica

 

Dear Carmen,

have you tried running ft_componentanalysis twice on the same matlab? The ICA algorithm has some intrinsic stochasticity in it, and therefore even running it twice on one machine and one matlab version should yield similar but not identical results.

I hope this is indeed the case, and that I'm not misleading you.

Best,

Roey

 

On Thu, Jul 16, 2015 at 11:05 AM, Carmen Kung <carmen.kung at mq.edu.au <mailto:carmen.kung at mq.edu.au> > wrote:

Dear fellow fieldtrippers,

A colleauge and I are analysing the same set of EEG data using the same fieldtrip toolbox (the version of 20150522) but we have different versions of matlab installed on each of our computer (mine is 2013b and hers is 2014b).  After running  ft_componentanalysis (using runica), we compared the output between us and we found that ft_componentanalysis yielded mostly the same components but still there are differences in ranking and the output.

Does anyone know why this is the case and suggest which Matlab version we should use

Many thanks in advance,
Carmen
_______________________________________________
fieldtrip mailing list
fieldtrip at donders.ru.nl <mailto:fieldtrip at donders.ru.nl> 
http://mailman.science.ru.nl/mailman/listinfo/fieldtrip

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20150716/8dff5056/attachment-0002.html>