-
Notifications
You must be signed in to change notification settings - Fork 1
/
emmv.html
68 lines (63 loc) · 3.43 KB
/
emmv.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE html>
<html>
<head>
<!-- <meta name="generator" content="jemdoc, see http://jemdoc.jaboc.net/" /> -->
<!-- <meta http-equiv="Content-Type" content="text/html;charset=UTF-8" /> -->
<link rel="stylesheet" href="jemdoc.css" type="text/css" />
<title>Nicolas Goix</title>
</head>
<body>
<script type="text/javascript">
var gaJsHost = (("https:" == document.location.protocol) ? "https://ssl." : "http://www.");
document.write(unescape("%3Cscript src='" + gaJsHost + "google-analytics.com/ga.js' type='text/javascript'%3E%3C/script%3E"));
</script>
<script type="text/javascript">
try {
var pageTracker = _gat._getTracker("UA-23998109-2");
pageTracker._trackPageview();
} catch(err) {}
</script>
<table summary="Table for page layout." id="tlayout">
<tr valign="top">
<td id="layout-menu">
<div class="menu-category">Nicolas Goix</div>
<div class="menu-item"><a href="index.html">Home</a></div>
<div class="menu-item"><a href="biography.html">Biography</a></div>
<div class="menu-category">Research</div>
<div class="menu-item"><a href="papers.html">Papers</a></div>
<div class="menu-category">Code </div>
<div class="menu-item"><a href="sklearn.html">Scikit-Learn</a></div>
<div class="menu-item"><a href="damex.html">Damex Algorithm</a></div>
<div class="menu-item"><a href="emmv.html" class="current">Unsupervised Evaluation</a></div>
<div class="menu-item"><a href="nyu.html">Black hole Cyg-X-1</a></div>
<div class="menu-item"><a href="ocrf.html">One Class Random Forests</a></div>
<div class="menu-category">Miscellaneous</div>
<div class="menu-item"><a href="links.html">Links</a></div>
</td>
<td id="layout-content">
<div id="toptitle">
<h1>Nicolas Goix – Evaluation of unsupervised Anomaly Detection algorithms</h1>
</div>
<!-- <h2>2011 – 2012</h2> -->
<ul>
<p>
<div class="infoblock">
When sufficient labeled data are available, classical criteria based on ROC or PR curves can be used to compare the performance of unsupervised anomaly detection algorithms. However, in many situations, few or no data are labeled. This calls for alternative criteria one can compute on non-labeled data. In
<a href="https://arxiv.org/abs/1607.01152" target="_blank"> [ICML 2016 workshop paper]</a>, two criteria that do not require labels are empirically shown to discriminate accurately (w.r.t. ROC or PR based criteria) between algorithms. These criteria are based on existing Excess-Mass (EM) and Mass-Volume (MV) curves, which generally cannot be well estimated in large dimension. A methodology based on feature sub-sampling and aggregating is also described and tested, extending the use of these criteria to high-dimensional datasets and solving major drawbacks inherent to standard EM and MV curves.
<p>
<p>
The associated code is available <a href="https://github.com/ngoix/EMMV_benchmarks" target="_blank"> here</a>.
</div>
<!-- <li><p>The implementation of DAMEX is available on my local version of scikit-learn, branch 'damex': [<a href="https://github.com/ngoix/scikit-learn/tree/damex" target="_blank">Code</a>].</p></li> -->
</ul>
<!-- <div id="footer"> -->
<!-- <div id="footer-text"> -->
<!-- Page generated by <a href="http://jemdoc.jaboc.net/">jemdoc</a>. -->
<!-- </div> -->
<!-- </div> -->
</td>
</tr>
</table>
</body>
</html>