Hello Mahesh,<div><br></div><div>See below in CAPS<br><br><div class="gmail_quote">Στις 6 Απριλίου 2011 5:51 μ.μ., ο χρήστης Mahesh Casiraghi <span dir="ltr"><<a href="mailto:mahesh.casiraghi@gmail.com">mahesh.casiraghi@gmail.com</a>></span> έγραψε:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;"><font><font><font><font><font><p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small">
<span><span style="font-size:10.5pt;font-family:Constantia;color:#5F5F5F">Hi Klados,</span></span></p>
<p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small"><span><span style="font-size:10.5pt;font-family:Constantia;color:#5F5F5F"><br></span></span></p><p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small">
<span><span style="font-size:10.5pt;font-family:Constantia;color:#5F5F5F">yep, the problem is running regression on components. It seems to me that for some reasons it tends to be generally more unstable than performing same algorithms on the source signal (chans x samps). Could be for the presumably bigger variance in magnitudes that we have between comonent and comoponent with respect to the one between chans and chans?</span></span></p>
<p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small"><span><span style="font-size:10.5pt;font-family:Constantia;color:#5F5F5F"><br></span></span></p></font></font></font></font></font></blockquote>
<div><br></div><div>YES THIS IS RIGHT IT IS MORE DIFFICULT TO RUN AN ADAPTIVE FILTER ON ICs RATHER THAN ON AFFECTED ELECTRODES THEMSELVES...ON THE OTHER HAND CLASSICAL REGRESSION APPROACHES ARE NOT AFFECTED BY THE VARIANCE IN MAGNITUDE...</div>
<div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;"><font><font><font><font><font><p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small">
<span><span style="font-size:10.5pt;font-family:Constantia;color:#5F5F5F"></span></span></p><p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small">
<span><span style="font-size:10.5pt;font-family:Constantia;color:#5F5F5F">And when one forces stability [SCRLS] on about 45 minutes [reshaped signal of interest from 3D (chans x samps x trials), 2d (chans x [samps,trials])] of 512Hz 69 chans EEG, using SCRLS_regression.m, it is practically impossible to keep numerical precision at 50 bits. </span></span></p>
</font></font></font></font></font></blockquote><div><br></div><div>45 MINS OF RECORDINGS WITH 512Hz SAMPLING FREQUENCY MEANS 45*60*512 SAMPLE POINTS... YOU CAN UNDERSTAND THAT THIS IS A HUGE SIGNAL FOR AN ADAPTIVE FILTER. I PROPOSE YOU TO BREAK DOWN THE PROCESS USING A LOOP BREAKING THE SIGNAL IN 1 MIN EPOCHS AND THEN FILTER IT WIT SCRLS </div>
<div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;"><font><font><font><font><font>
<p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small"><span><span style="font-size:10.5pt;font-family:Constantia;color:#5F5F5F"><br></span></span></p><p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small">
<span><span style="font-size:10.5pt;font-family:Constantia;color:#5F5F5F">I had to go down to 30-34 bits (currently trying to do that at 30, which is sensibly faster than 32 and 34), and I was wondering how this trade off in numerical precision will in effect affect the quality of the regression, and if and how playing with sigma and labmda may produce better results.</span></span></p>
<p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small"><span><span style="font-size:10.5pt;font-family:Constantia;color:#5F5F5F"><br></span></span></p></font></font></font></font></font></blockquote>
<div><br></div><div>THIS IS A VERY GOOD QUESTION FOR RESEARCH... :)</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;"><font><font><font><font><font><p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small">
<span><span style="font-size:10.5pt;font-family:Constantia;color:#5F5F5F"></span></span></p><p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small">
<span><span style="font-size:10.5pt;font-family:Constantia;color:#5F5F5F">Thank you for your skype contact, I am going to run some more tests and I will contact you with more impressions.</span></span></p><p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small">
<span><span style="font-size:10.5pt;font-family:Constantia;color:#5F5F5F"><br></span></span></p></font></font></font></font></font></blockquote><div><br></div><div>I AM GLAD TO HELP YOU AS WELL AS THE WHOLE RESEARCH COMMUNITY IF IT IS POSSIBLE</div>
<div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;"><font><font><font><font><font><p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small">
<span><span style="font-size:10.5pt;font-family:Constantia;color:#5F5F5F"></span></span></p><font color="#888888"><p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px">Mahesh</span></font></p>
</font><p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px"><br></span></font></p><p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px">ps> Of course I was meaning Klados, I think I got confused by some previous messages.</span></font></p>
<p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small"><span><span style="font-size:10.5pt;font-family:Constantia;color:#5F5F5F"><br></span></span></p><p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small">
<span><span style="font-size:10.5pt;font-family:Constantia;color:#5F5F5F"> </span></span></p></font></font></font>
<br clear="all"></font></font><div class="im"><div style="text-align:right"><span style="font-family:verdana, sans-serif;font-size:x-small;color:rgb(153, 153, 153)"><br></span></div><div style="text-align:right"><span style="font-family:verdana, sans-serif;font-size:x-small"><font color="#C0C0C0"><br>
</font></span></div><div style="text-align:right"><span style="font-family:verdana, sans-serif;font-size:x-small"><font color="#C0C0C0"><br></font></span></div><div style="text-align:right"><span style="border-collapse:collapse"><div style="font-size:13px;font-family:arial, sans-serif">
<font><span style="font-size:x-small"><font face="verdana, sans-serif" color="#999999">Mahesh M. Casiraghi</font></span></font></div><div style="font-size:13px;font-family:arial, sans-serif"><span style="font-size:x-small"><font face="verdana, sans-serif" color="#999999">PhD candidate - Cognitive Sciences</font></span></div>
<div><font size="1" face="verdana, sans-serif" color="#999999">Roberto Dell'Acqua Lab, University of Padova</font></div><div><font size="1" face="verdana, sans-serif" color="#999999">Pierre Jolicoeur Lab, Univesité de Montréal</font></div>
<div style="color:rgb(192, 192, 192)"><font size="1" face="verdana, sans-serif"><a href="mailto:mahesh.casiraghi@umontreal.ca" target="_blank">mahesh.casiraghi@umontreal.ca</a></font></div></span></div><div style="text-align:right">
<font size="1" face="verdana, sans-serif" color="#C0C0C0"><br></font></div><div style="text-align:right"><font size="1" face="verdana, sans-serif" color="#999999">I have the conviction that when Physiology will be far enough advanced, the poet, the philosopher, and the physiologist will all understand each other.</font></div>
<div style="text-align:right"><font size="1" face="verdana, sans-serif" color="#999999">Claude Bernard</font></div><div style="text-align:right"><font color="#CCCCCC" face="constantia, 'hoefler text', 'palatino linotype', serif" size="1"><span style="border-collapse:collapse;line-height:24px"><br>
</span></font></div><br>
<br><br></div><div><div></div><div class="h5"><div class="gmail_quote">2011/4/5 Μανούσος Κλάδος <span dir="ltr"><<a href="mailto:mklados@gmail.com" target="_blank">mklados@gmail.com</a>></span><br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div bgcolor="#FFFFFF"><div>Dear Mahesh,</div><div><br></div><div>Let me understand something. In this kind of dataset ICA is running and the problem is with the regression part? If </div><div>so you can perform reg in smaller segments of ics... Also you are able to try another regression scheme based one a single step regression and not in adaptive filters...</div>
<div><br></div><div>Also watch your eog signals ... You have 4 channels but from these channels 2 bipolar signals<span> are obtained which are going to be used as an input in every adaptive filter...</span></div><div><span><br>
</span></div><div><span>Everything you need About regica I am available for all of you in my e-mail or even better at Skype(mklados) you can also follow me at twitter (@mklados) where future versions of regica will be announced as well as many other artifact rejection and neuroscientific staff will be pointed</span></div>
<div><span><br></span></div><div><span>Sincerely yours</span></div><div><span>Manousos Klados</span></div><div><span><br></span></div><div><span>P.S.</span></div><div><span><br></span></div><div><span>Which is Florian?</span></div>
<div><span><br></span></div><div><span><br></span></div><div>___________________________<div>Manousos Klados<br><div>PhD Candidate</div><div>Group of Applied Neuroscience </div><div>Lab of Medical Informatics</div><div>Medical School</div>
<div>Aristotle University of Thessaloniki</div><div>___________________________<br><div>iPhone</div></div></div></div><div><br>4 Απρ 2011, 11:10 μ.μ., ο/η Mahesh Casiraghi <<a href="mailto:mahesh.casiraghi@gmail.com" target="_blank">mahesh.casiraghi@gmail.com</a>> έγραψε:<br>
<br></div><div><div></div><div><div></div><blockquote type="cite"><div><font><font><font><font><font><p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small"><span><span style="font-size:10.5pt;font-family:Constantia;color:#5F5F5F">Dear EEGLabbers,</span></span></p>
<p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small"><span><span style="font-size:10.5pt;font-family:Constantia;color:#5F5F5F"><br></span></span></p><p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small">
<span><span style="font-size:10.5pt;font-family:Constantia;color:#5F5F5F">I am trying here to test if the hybrid methodology proposed by Florian and his group [</span></span><span style="color:rgb(255, 255, 255);font-family:verdana, sans-serif"><a href="http://lomiweb.med.auth.gr/gan/mklados/index.php?option=com_k2&view=item&id=25:regica" target="_blank"></a><a href="http://lomiweb.med.auth.gr/gan/mklados/index.php?option=com_k2&view=item&id=25:regica" target="_blank">http://lomiweb.med.auth.gr/gan/mklados/index.php?option=com_k2&view=item&id=25:regica</a></span><span style="color:rgb(95, 95, 95);font-family:Constantia;font-size:14px">] may be effective in removing eye-artifacts in an experiment where 10 secs epochs need to be segmented.</span></p>
<p class="MsoNormal" style="color:rgb(255, 255, 255);font-family:verdana, sans-serif;font-size:small"><span style="color:rgb(95, 95, 95);font-family:Constantia;font-size:14px"><br></span></p>
<p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px">The concept of the methodology seems promising to me, but I am nonetheless a bit puzzled with respect to which regression algorithm might be adopted. When it comes to run the code on a real subject (512Hz sampling rate, continuous data, 64 EEG and 4 EOG chans, about 1.30 hours) LMS and CRLS become both unstable and fail after just few steps. H INF ew and tv algorithms, do that too. It seems to me the only option left is to make use of SCRLS_regression.m from the AAR toolbox, but as the relative documentation suggests, the function is not really optimized for fast computation, and the reg procedure seems to take ages to converge [running it in matlab 64, on a 4cores pc, for a 1.30 hours continuous EEG and it is still trying to converge after 23 hours of computation, just one processor used].</span></font></p>
<p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px"><br></span></font></p><p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px">Question is: can someone out there with some experience with AAR toolbox and/or SCRLS algorithm provide some insights on how to play around with the 'lambda', 'sigma', and 'precision' fields of the opt structure so as to come up with a sufficiently accurate output in an acceptable amount of time? I was unable to find any detailed summary or list of practical guidelines/hints concerning these parameters. Furthermore, perhaps someone is aware of a more effective SRLS reg routine...</span></font></p>
<p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px"><br></span></font></p><p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px">Here the code I used to reshape the tri-dimensional chans x samps x trials matrix, run regica, and come back to the cleaned 3 dims matrix. As you see, opt parameters are default, with the exception of .20 instead of .25 for correlation threshold, note that srls is default here. </span></font></p>
<p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px"><br></span></font></p><p class="MsoNormal"><font color="#5F5F5F" face="Constantia"></font></p>
<p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px"> EEG2DIM.data = reshape(EEG.data, size(EEG.data,1),size(EEG.data,3)*size(EEG.data,2));</span></font></p>
<p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px"> opt.EOG = [EEG2DIM.data(1,:);EEG2DIM.data((67:69),:)]; %Fp1(1) plus HEOG1, HEOG2, & VEOG</span></font></p>
<p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px"> opt.M=3;</span></font></p><p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px"> opt.lambda=0.9999;</span></font></p>
<p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px"> opt.sigma=0.01;</span></font></p><p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px"> opt.prec=50;</span></font></p>
<p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px"> opt.crittype = 'correlation';</span></font></p><p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px"> opt.corthr = 20;</span></font></p>
<p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px"> [EEG1] = regica(EEG2DIM.data((1:64),:),opt);</span></font></p><p class="MsoNormal">
<font color="#5F5F5F" face="Constantia"><span style="font-size:14px"> [EEG2] = reshape(EEG1, size(EEG1,1), size(EEG.data,2)/size(EEG.data,3), size(EEG.data,3));</span></font></p>
<font color="#5F5F5F" face="Constantia"><span style="font-size:14px"> EEG.data((1:64),:) = EEG2((1:64),:);</span></font><p></p><p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px"><br>
</span></font></p><p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px">Any insight would be really appreciated,</span></font></p><p class="MsoNormal">
<font color="#5F5F5F" face="Constantia"><span style="font-size:14px"><br></span></font></p><p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px">Mahesh</span></font></p>
<p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px"><br></span></font></p><p class="MsoNormal"><font color="#5F5F5F" face="Constantia"><span style="font-size:14px"> </span></font></p>
</font></font></font></font></font><div style="text-align:right"><span style="font-family:verdana, sans-serif;font-size:x-small"><font color="#C0C0C0"><br></font></span></div><div style="text-align:right"><span style="font-family:verdana, sans-serif;font-size:x-small"><font color="#C0C0C0"><br>
</font></span></div><div style="text-align:right"><span style="border-collapse:collapse"><div style="font-size:13px;font-family:arial, sans-serif"><font><span style="font-size:x-small"><font face="verdana, sans-serif" color="#999999">Mahesh M. Casiraghi</font></span></font></div>
<div style="font-size:13px;font-family:arial, sans-serif"><span style="font-size:x-small"><font face="verdana, sans-serif" color="#999999">PhD candidate - Cognitive Sciences</font></span></div><div><font size="1" face="verdana, sans-serif" color="#999999">Roberto Dell'Acqua Lab, University of Padova</font></div>
<div><font size="1" face="verdana, sans-serif" color="#999999">Pierre Jolicoeur Lab, Univesité de Montréal</font></div><div style="color:rgb(192, 192, 192)"><font size="1" face="verdana, sans-serif"><a href="mailto:mahesh.casiraghi@umontreal.ca" target="_blank"></a><a href="mailto:mahesh.casiraghi@umontreal.ca" target="_blank">mahesh.casiraghi@umontreal.ca</a></font></div>
</span></div><div style="text-align:right"><font size="1" face="verdana, sans-serif" color="#C0C0C0"><br></font></div><div style="text-align:right"><font size="1" face="verdana, sans-serif" color="#999999">I have the conviction that when Physiology will be far enough advanced, the poet, the philosopher, and the physiologist will all understand each other.</font></div>
<div style="text-align:right"><font size="1" face="verdana, sans-serif" color="#999999">Claude Bernard</font></div><div style="text-align:right"><font color="#CCCCCC" face="constantia, 'hoefler text', 'palatino linotype', serif" size="1"><span style="border-collapse:collapse;line-height:24px"><br>
</span></font></div><br>
</div></blockquote></div></div><blockquote type="cite"><div><span>_______________________________________________</span><br><span>Eeglablist page: <a href="http://sccn.ucsd.edu/eeglab/eeglabmail.html" target="_blank"></a><a href="http://sccn.ucsd.edu/eeglab/eeglabmail.html" target="_blank">http://sccn.ucsd.edu/eeglab/eeglabmail.html</a></span><br>
<span>To unsubscribe, send an empty email to <a href="mailto:eeglablist-unsubscribe@sccn.ucsd.edu" target="_blank"></a><a href="mailto:eeglablist-unsubscribe@sccn.ucsd.edu" target="_blank">eeglablist-unsubscribe@sccn.ucsd.edu</a></span><br>
<span>For digest mode, send an email with the subject "set digest mime" to <a href="mailto:eeglablist-request@sccn.ucsd.edu" target="_blank"></a><a href="mailto:eeglablist-request@sccn.ucsd.edu" target="_blank">eeglablist-request@sccn.ucsd.edu</a></span></div>
</blockquote></div></blockquote></div><br>
</div></div></blockquote></div><br><br clear="all"><br>-- <br>Manousos A. Klados<br>PhD Candidate -- Research Assistant<br>Group of Applied Neurosciences<br>Lab of Medical Informatics<br>School of Medicine<br>Aristotle University of Thessaloniki<br>
P.O. Box 323 54124 Thessaloniki Greece<br>_________________________________________________<br>Tel: +30-2310-999332<br>Fax:+30-2310-999263<br>Website: <a href="http://lomiweb.med.auth.gr/gan/mklados">http://lomiweb.med.auth.gr/gan/mklados</a><br>
<br>________________________________________________________________<br>Δρώ γιατί Αντιδρώ: Δεν τυπώνω αυτό το mail γιατί προστατεύω το περιβάλλον.<br>Acting by Reacting: By not printing this e-mail I help protect the environment.<br>
________________________________________________________________<br>
</div>