Dear ICA persons and list,<div><br></div><div>Mo observed FastICA showed more mutual information reduction than Infomax. I'm curious to know if FastICA outperforms Infomax under some conditions. Any comment?</div><div>
<br></div><div><div>Makoto</div><div><br><div class="gmail_quote">2013/3/19 Mo Khalili <span dir="ltr"><<a href="mailto:mokhal@yahoo.co.uk" target="_blank">mokhal@yahoo.co.uk</a>></span><br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div><div style="font-size:12pt;font-family:times new roman,new york,times,serif"><div style="font-family:'times new roman','new york',times,serif;font-size:12pt"><span>Dear Makato,</span></div><div style="font-style:normal;font-size:16px;background-color:transparent;font-family:'times new roman','new york',times,serif">
<span><br></span></div><div style="background-color:transparent"><span>I am using the freely available dataset of the same paper (</span><span style="font-style:normal;font-size:12pt;font-family:'times new roman','new york',times,serif">Delorme et al.(2012) in PLoS One). Full details of the experimental paradigm and dataset is provided in the following link: <a href="http://sccn.ucsd.edu/wiki/BSS_Comparison" target="_blank">http://sccn.ucsd.edu/wiki/BSS_Comparison</a>.</span></div>
<div><span style="font-style:normal;font-size:12pt;font-family:'times new roman','new york',times,serif">Actually I only calculated the averaged absolute correlation between estimated sources.</span></div>
<div style="font-style:normal;font-size:12pt;background-color:transparent;font-family:'times new roman','new york',times,serif"><span style="font-style:normal;font-size:12pt;font-family:'times new roman','new york',times,serif"><br>
</span></div><div style="font-style:normal;font-size:16px;background-color:transparent;font-family:'times new roman','new york',times,serif"><span style="font-style:normal;font-size:12pt;font-family:'times new roman','new york',times,serif">Thanks</span></div>
<div style="font-style:normal;font-size:16px;background-color:transparent;font-family:'times new roman','new york',times,serif">Mo</div><div style="font-style:normal;font-size:16px;background-color:transparent;font-family:'times new roman','new york',times,serif">
<br></div> <div style="font-family:'times new roman','new york',times,serif;font-size:12pt"> <div style="font-family:'times new roman','new york',times,serif;font-size:12pt"> <div dir="ltr">
<font face="Arial"><div class="im"> <hr size="1"> <b><span style="font-weight:bold">From:</span></b> Makoto Miyakoshi <<a href="mailto:mmiyakoshi@ucsd.edu" target="_blank">mmiyakoshi@ucsd.edu</a>><br> <b><span style="font-weight:bold">To:</span></b> Mo Khalili <<a href="mailto:mokhal@yahoo.co.uk" target="_blank">mokhal@yahoo.co.uk</a>> <br>
</div><b><span style="font-weight:bold">Cc:</span></b> EEGLAB List <<a href="mailto:eeglablist@sccn.ucsd.edu" target="_blank">eeglablist@sccn.ucsd.edu</a>> <br> <b><span style="font-weight:bold">Sent:</span></b> Tuesday, 19 March 2013, 16:51<div>
<div class="h5"><br> <b><span style="font-weight:bold">Subject:</span></b> Re:
[Eeglablist] paradox in whitening<br> </div></div></font> </div><div><div class="h5"> <br><div>Dear Mo,<div><br></div><div>Ok, why don't you tell us about your data (number of channels, data length, task, recording system, etc). We may be able to find a reason.</div>
<div><br></div><div>Makoto<br><br><div>
2013/3/18 Mo Khalili <span dir="ltr"><<a rel="nofollow" href="mailto:mokhal@yahoo.co.uk" target="_blank">mokhal@yahoo.co.uk</a>></span><br><blockquote style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div><div style="font-size:12pt;font-family:'times new roman','new york',times,serif"><div style="font-size:12pt">Hi Makoto,</div><div style="font-size:12pt">
<br></div><div style="background-color:transparent">yes, I used the same dataset with multiple runs of ICA (different decompositions); there wasn't any significant change in the results.</div><div style="background-color:transparent">
the absolute correlation between pairs of components ,resulting from FastICA, is near zero, so the components are well decorrelated. However, in the infomax the result shows a significantly higher correlation between IC pairs.</div>
<span><font color="#888888"><div style="font-size:12pt"><br></div><div style="font-style:normal;font-size:16px;background-color:transparent">
Mo</div> </font></span><div style="font-size:12pt"> <div style="font-size:12pt"> <div dir="ltr">
<font face="Arial"> <hr size="1"><div> <b><span style="font-weight:bold">From:</span></b> Makoto Miyakoshi <<a rel="nofollow" href="mailto:mmiyakoshi@ucsd.edu" target="_blank">mmiyakoshi@ucsd.edu</a>><br> <b><span style="font-weight:bold">To:</span></b> Mo Khalili <<a rel="nofollow" href="mailto:mokhal@yahoo.co.uk" target="_blank">mokhal@yahoo.co.uk</a>> <br>
<b><span style="font-weight:bold">Cc:</span></b> "<a rel="nofollow" href="mailto:eeglablist@sccn.ucsd.edu" target="_blank">eeglablist@sccn.ucsd.edu</a>" <<a rel="nofollow" href="mailto:eeglablist@sccn.ucsd.edu" target="_blank">eeglablist@sccn.ucsd.edu</a>> <br>
<b><span style="font-weight:bold">Sent:</span></b> Tuesday, 19 March 2013, 0:23<br> <b><span style="font-weight:bold">Subject:</span></b> Re: [Eeglablist] paradox in whitening<br> </div></font><div> </div>
</div><div><div> <br><div>Dear
Mo,<div><br></div><div><font color="#222222" face="arial, sans-serif">That's not consistent with Delorme et al. (2012) in PLoS One. I wonder why too. Did you use the same data to test two different weight matrices?</font></div>
<div><font color="#222222" face="arial, sans-serif"><br></font></div><div><font color="#222222" face="arial, sans-serif">Makoto</font></div><div><br></div><div><br><br><div>2013/3/18 Mo Khalili <span dir="ltr"><<a rel="nofollow" href="mailto:mokhal@yahoo.co.uk" target="_blank">mokhal@yahoo.co.uk</a>></span><br>
<blockquote style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div><div style="font-size:12pt"><div><span style="font-family:arial,sans-serif;color:rgb(34,34,34);font-size:13px">Hi,</span><br>
</div><div style="font-size:12pt"><div style="font-size:12pt"><div><div><div style="font-size:12pt">
<div style="font-size:12pt"><div style="font-size:12pt"><div><div style="font-size:12pt">
<div style="font-family:arial,sans-serif;font-size:13px;color:rgb(34,34,34)"><br></div><div style="font-family:arial,sans-serif;font-size:13px;color:rgb(34,34,34)">I have a general query about performance of ICA algorithms.</div>
<div><font color="#222222" face="arial, sans-serif">As I know FastICA is trying to estimate sources based
on maximizing the non-gaussianity while, infoamx is trying to estimate the components based on minimizing mutual information between sources. What wonders me is , when I calculated the mutual information between ICs, the averaged mutual information between pairs of ICs in FastICA is lower than ICs estimated by infomax. Additionally when I calculate
the correlation between pairs of the estimated ICs, the averaged absolute correlation between ICs obtained by FastICA is significantly lower than infomax.</font></div><div><font color="#222222" face="arial, sans-serif">in my opinion this result speculates, the whitening(sphering) step in the infomax is different from FastICA, and it does not perform the same. </font></div>
<div><font color="#222222" face="arial, sans-serif">However, when I look at the codes of FastICA and infomax, both have used the same method (and functions) for whitening.</font></div><div style="font-family:arial,sans-serif;font-size:13px;color:rgb(34,34,34)">
I appreciate it if anyone can help me through understanding the reason.</div><div style="font-family:arial,sans-serif;font-size:13px;color:rgb(34,34,34)"><br></div><div style="font-family:arial,sans-serif;font-size:13px;color:rgb(34,34,34)">
Regards,</div><div style="font-family:arial,sans-serif;font-size:13px;color:rgb(34,34,34)">Mo</div></div></div><br><br> </div> </div> </div></div></div><br><br> </div> </div> </div></div><br>_______________________________________________<br>
Eeglablist page: <a rel="nofollow" href="http://sccn.ucsd.edu/eeglab/eeglabmail.html" target="_blank">http://sccn.ucsd.edu/eeglab/eeglabmail.html</a><br>
To unsubscribe, send an empty email to <a rel="nofollow" href="mailto:eeglablist-unsubscribe@sccn.ucsd.edu" target="_blank">eeglablist-unsubscribe@sccn.ucsd.edu</a><br>
For digest mode, send an email with the subject "set digest mime" to <a rel="nofollow" href="mailto:eeglablist-request@sccn.ucsd.edu" target="_blank">eeglablist-request@sccn.ucsd.edu</a><br></blockquote></div><br>
<br clear="all"><div><br></div>
-- <br>Makoto Miyakoshi<br>JSPS Postdoctral Fellow for Research Abroad<br>Swartz Center for Computational Neuroscience<br>Institute for Neural Computation, University of California San Diego<br>
</div>
</div><br><br> </div></div></div> </div> </div></div></blockquote></div><br><br clear="all"><div><br></div>-- <br>Makoto Miyakoshi<br>JSPS Postdoctral Fellow for Research Abroad<br>Swartz Center for Computational Neuroscience<br>
Institute for Neural Computation, University of California San Diego<br>
</div>
</div><br><br> </div></div></div> </div> </div></div></blockquote></div><br><br clear="all"><div><br></div>-- <br>Makoto Miyakoshi<br>JSPS Postdoctral Fellow for Research Abroad<br>Swartz Center for Computational Neuroscience<br>
Institute for Neural Computation, University of California San Diego<br>
</div></div>