<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta http-equiv="Content-Type" content="text/html;charset=ISO-8859-1">
<title></title>
</head>
<body text="#000000" bgcolor="#ffffff">
Dear Jim,<br>
<blockquote type="cite"
cite="mid6.1.2.0.2.20050304131234.01c4b458@pop.nmsu.edu"><font size="3">Hello.
We wish to know more about the algorithm used
by EEGLAB to do ICA. Looking at the reference list for
"EEGLAB: an open source toolbox for analysis of single-trial EEG
dynamics including independent component analysis," it seems the
best sources are the chapters in "Advances in Neural Information
Processing Systems," which we unfortunately don't have immediate
access to. </font></blockquote>
There are a number of ressources on the Internet about ICA (some of
them listed below). <br>
<br>
<a class="moz-txt-link-freetext" href="http://sccn.ucsd.edu/~arno/indexica.html">http://sccn.ucsd.edu/~arno/indexica.html</a> (no equations)<br>
<a class="moz-txt-link-freetext" href="http://www.cis.hut.fi/aapo/papers/IJCNN99_tutorialweb/">http://www.cis.hut.fi/aapo/papers/IJCNN99_tutorialweb/</a> (good reference)<br>
<a class="moz-txt-link-freetext" href="http://www.ph.tn.tudelft.nl/~dick/cvonline/ica/ica.html">http://www.ph.tn.tudelft.nl/~dick/cvonline/ica/ica.html</a> <br>
<a class="moz-txt-link-freetext" href="http://www.fmrib.ox.ac.uk/analysis/research/melodic/tr02cb1/node1.html">http://www.fmrib.ox.ac.uk/analysis/research/melodic/tr02cb1/node1.html</a>
(technical but very detailed for application to fMRI)<br>
<blockquote type="cite"
cite="mid6.1.2.0.2.20050304131234.01c4b458@pop.nmsu.edu"><font size="3">So,
we wonder what references would be recommended to
understand the algorithm for ICA, as well as the Matlab implementation
of
it.<br>
</font></blockquote>
About Infomax ICA implemented in EEGLAB, the best reference is probably
Tony Bell paper<br>
<br>
<big><small>Bell AJ, Sejnowski TJ.<br>
An information-maximization approach to blind separation and blind
deconvolution.</small><br>
<font size="-1"><big><span
onmouseover="AbbrLookUp(this, 'Neural Comput.');">Neural Comput.</span>
1995 Nov;7(6):1129-59. </big></font><br>
<br>
<font size="-1"><big>For the Matlab implementation, the best is
probably to look at the Matlab code itself (runica.m). There are other
implementation of Infomax (Natural Gradient) in the ICALAB toolbox
(which is automatically detected by EEGLAB).</big></font></big><font
size="-1"><br>
</font>
<blockquote type="cite"
cite="mid6.1.2.0.2.20050304131234.01c4b458@pop.nmsu.edu"><font size="3">Particularly,
we wonder if a mixing matrix that maps components to
electrodes is computed during the calculations, whether it is
separately
computed for each "trial," and how we can access this matrix in
Matlab.<br>
</font></blockquote>
Yes, the mixing matrix is computed at each time step. It is not
computed separately for each trial. All data points from all trials are
shuffled at each time step. Infomax ICA does not use the fact that
neighboring time points have similar activities (other ICA algorithms
like SOBI do, but it does not mean that they are returning better
results). <br>
<br>
Best,<br>
<br>
Arno<br>
<br>
</body>
</html>