SIMILARITY KERNELS FOR DISTANCE-BASED OUTLIER DETECTION Ruben Ramirez Padron, PhD candidate Department of Electrical Engineering and Computer Science University of Central Florida FEBRUARY 25, 2011 10:30AM – 11:30PM Key West Ballroom 218D Abstract Outlier detection is an important research topic that focuses on detecting abnormal information in data sets and processes. This talk will address the problem of determining which class of kernels should be used in a geometric framework for distance-based outlier detection. A new class of similarity kernels is proposed to tackle this problem. Additionally, a new similarity score is introduced: the summation kernel similarity score (SKSS)*.* Preliminary experimental results showed that SKSS compare favorably to the commonly used k-NN kernel similarity score (kNNSS). BIO: Ruben Ramirez Padron graduated with a Master degree in Computer Engineering from UCF in 2009. He graduated with a Bachelor degree in Computer Science from "Universidad Central de Las Villas", Cuba, in 1996. Currently, he is a PhD candidate at the Department of Electrical Engineering and Computer Science at UCF, under the guidance of Dr. Avelino Gonzalez. Ruben's main research interests are: novelty detection, kernel methods for pattern analysis, and online machine learning.