Computational Neurobiology Lab

Sigurd Enghoff

The following are links to source code and binary directories for stand-alone and MEX-file implementations of the MATLAB runica.m routine available in the EEGLAB Toolbox for MATLAB.

The ICA program implements the infomax and extended-infomax algorithms of Bell and Sejnowski (1995) and Lee, Girolami and Sejnowski (1998). The MEX version was written under MATLAB version 5.2. The stand-alone ICA program analyzes data stored as a (native machine) float matrix (multiplexed by channel). It requires an ASCII script file (see example in "ica.sc").

Makefiles and executables are available for DEC Alpha, SGI and Sun 4. The SGI executables require the BLAS library to run and compile. Most SGI systems have this library available. Recompiling either program requires the BLAS and LAPACK libraries. Precompiled versions are available below.

The sample stand-alone script file "ica.sc" uses the sample EEG data file "floatdata." Native-float versions of this file are provided. The programs can be applied to a very wide range of applications.

Further information about ICA is available from the ICA-CNL web pages. Also, please consult my thesis Moving ICA and Time-Frequency Analysis in Event-Related EEG Studies of Selective Attention.

All software contained on this site is covered by the GNU General Public License.