<html>
<head>
<meta content="text/html; charset=utf-8" http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<div align="center"><font size="+2">The ICLabel Website -- A New
Education Feature</font></div>
<p>I want to update everyone on a new education feature available at
<a href="reaching.ucsd.edu:8000" moz-do-not-send="true">the
ICLabel website</a> (<a moz-do-not-send="true"
href="reaching.ucsd.edu:8000">reaching.ucsd.edu:8000</a>). The
site is designed to collect 'crowdsourced' labels for EEG
independent components (ICs), based on their displayed properties.
I am using the submitted labels to train an automated IC
classifier However, at the recent EEGLAB workshop the site also
gained a fair bit of favorable response as a useful educational
tool using the new option described below.</p>
<div bgcolor="#FFFFFF">
<p><b>For new users wanting to learn the basics about classifying
EEG ICs based on their measured properties: </b> I have
created a <a moz-do-not-send="true"
href="reaching.ucsd.edu:8000/labelfeedback">Label with
feedback</a> option. This page is almost exactly the original
labeling page, but with three key differences:</p>
<ol>
<li>Feedback is provided after each IC label submission.<br>
</li>
<li>Submitted labels are not saved in the classification
database.</li>
<li>No login is required to reach the page.</li>
</ol>
The feedback provided to users is a table of all other labels for
that IC given by other people. Each label is marked as either by a
"user" who has signed up through the website or by an "expert"
(e.g., a high-skill user known to me). To clarify: "user" here
does not mean "worse than expert", but simply a volunteer label
contributor whose skill level is unknown.<br>
</div>
<div><br>
</div>
<div>There is still a <a moz-do-not-send="true"
href="reaching.ucsd.edu:8000/tutorial">Tutorial</a> section
giving sample classifications with text explanations. The <a
moz-do-not-send="true"
href="reaching.ucsd.edu:8000/labelfeedback">Label with feedback</a>
feature is suitable for learning and/or teaching by any researcher
or research student who wants to learn more about applying ICA
decomposition to EEG or related data. <br>
<br>
</div>
<div bgcolor="#FFFFFF"><b>For (self-selected) contributors to the
crowdsourced label database:</b> A second new feature is the
user Profile page that now appears by default following login. It
currently does not have very much information on your previous
contributions, but that will change with time -- I plan to
incorporate the output of a crowd-labelling algorithm that
estimates users' skill by comparing submitted labels with those of
established 'expert' users (i.e. those whose labels are most
consistent with other skilled contributors). If there is anything
else you would find useful to see there, do not hesitate to tell
me, either by replying to this email or by leaving a comment on
the website.<br>
<br>
I hope you all find the new <a moz-do-not-send="true"
href="reaching.ucsd.edu:8000/labelfeedback">Label with feedback</a>
page to be helpful and, after using it, some more of you will
choose to become contributors who <a moz-do-not-send="true">submit
more IC labels</a> to help the classification algorithm become
more accurate. </div>
<div> </div>
Finally, I have some working prototypes of the classifier and plan
to post a sort of progress report on the website, but I will leave
that for a future email.<br>
<br>
Sincerely,<br>
Luca Pion-Tonachini<br>
<p><br>
</p>
</body>
</html>