Back to the main page.
Bug 2256 - ft_componentanalysis with method = 'binica' does not work, also consider implementing 'mexica'?
| Status | CLOSED WONTFIX |
| Reported | 2013-08-19 16:10:00 +0200 |
| Modified | 2014-01-29 13:28:40 +0100 |
| Product: | FieldTrip |
| Component: | core |
| Version: | unspecified |
| Hardware: | PC |
| Operating System: | Windows |
| Importance: | P3 normal |
| Assigned to: | Eelke Spaak |
| URL: | |
| Tags: | |
| Depends on: | |
| Blocks: | |
| See also: |
Eelke Spaak - 2013-08-19 16:10:47 +0200
When I try to run binica, I get this message: Warning: adding /home/electromag/eelspa/ft-src/external/eeglab toolbox to your Matlab path Undefined function or variable 'icadefs'. Error in binica (line 95) icadefs % import ICABINARY and SC Error in ft_componentanalysis (line 480) [weights, sphere] = binica(dat, optarg{:}); Apparently some file from eeglab is missing. Apart from this, there also is a MEX-interface to EEGlab's ICA routines (see http://sccn.ucsd.edu/~scott/tutorial/ica.help). Should we consider making this accessible through ft_componentanalysis? The default runica is very slow; EEGlab documentation suggests that binica is up to 12x faster. (I guess this holds to a certain extent for mexica as well. binica requires a lot of disk I/O, as the data is written to a file and an executable is ran to process that file. Therefore I think mexica is a better idea for raw data.)
Jörn M. Horschig - 2013-08-19 16:19:18 +0200
mexica, lol... is there also corsica and veronica?
Eelke Spaak - 2013-08-26 11:15:40 +0200
It seems that binica is actually slower than runica, at least on our cluster... Maybe this is because our torque sessions can only use one cpu per session, while binica supports multicore computation? For now, marking this as wontfix, not really worthwhile.