<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
<title></title>
</head>
<body bgcolor="#ffffff" text="#000000">
On 08/26/2010 05:01 PM, Karch, Julian wrote:
LibSVM Support Vectors
<!-- Converted from text/plain format --><br>
<blockquote
cite="mid:82EE999347EBF142A8D78EF16FB83D8D01C824D9@MPIBMAIL01.mpib-berlin.mpg.de"
type="cite">
<p><font size="2">can somebody tell me how i get the support vectors
of a SVM instance?<br>
</font></p>
</blockquote>
<font size="2"><br>
Hi,<br>
<br>
in PyMVPA '0.5.0.dev' [0] you can get support vectors from a trained SVM<br>
instance (assume we call it clf) with: clf.model.get_sv()<br>
Note that clf.model has other useful methods involving support vectors.<br>
<br>
Prior to </font><font size="2">'0.5.0.dev' I guess there is no easy
way to get them.<br>
<br>
Best,<br>
<br>
Emanuele<br>
<br>
[0]: this version is the master branch of the git repository<br>
<a class="moz-txt-link-freetext" href="http://github.com/hanke/PyMVPA">http://github.com/hanke/PyMVPA</a><br>
which you can get via 'git clone...' or just clicking "download<br>
source" there.<br>
A nice tutorial related to this version is here:<br>
<a class="moz-txt-link-freetext" href="http://dev.pymvpa.org/tutorial.html">http://dev.pymvpa.org/tutorial.html</a><br>
</font>
</body>
</html>