BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Stochastic Outlier Selection - Jeroen Janssens (Tilburg University
   / University of Cambridge)
DTSTART:20100308T160000Z
DTEND:20100308T163000Z
UID:TALK23645@talks.cam.ac.uk
CONTACT:Zoubin Ghahramani
DESCRIPTION:Anomaly detection is relevant for many tasks\, ranging from de
 tecting credit card fraud to terrorist threats. Typically\, outlier-select
 ion algorithms are used for detecting anomalies. First\, we briefly explai
 n the difference between anomalies and outliers. Subsequently\, we present
  a novel\, unsupervised outlier-selection algorithm\, called Stochastic Ou
 tlier Selection (SOS). The SOS algorithm computes for each data point an o
 utlier probability. These probabilities are much more intuitive than the u
 nbounded outlier scores computed by existing outlier-selection algorithms.
  We evaluate SOS on a variety of real-world and synthetic datasets\, and c
 ompare it to four state-of-the-art outlier-selection algorithms. Our resul
 ts show that SOS has a superior performance while being more robust to dat
 a perturbations and parameter settings. We conclude that SOS is an effecti
 ve algorithm to select outliers in a dataset that compares favorably to st
 ate-of-the-art outlier-selection algorithms.\n\nThis is work with Eric O. 
 Postma.\n\nhttp://www.jeroenjanssens.com
LOCATION:Engineering Department\, CBL Room 438
END:VEVENT
END:VCALENDAR