"Learning and Classification in Networked Data"

11/10/05: 10:30 AM, webcast
11th Floor Large Conference Room
Host: Patrick Pantel, schedule

Abstract: Networked objects such as web-pages, research papers, people and companies form a complex structure of interdependence. In this talk, I explore classifying such objects using only the network structure and ignoring intrinsic attributes of the objects. Networked objects differ from the traditional machine learning point of view where instances are independent. Networked objects clearly are not. Further, networked data has the property that often the instances learned from are connected to instances whose class is to be estimated. Therefore, known instances has dual roles, first as training examples and later as background knowledge. I will introduce NetKit, a network learning toolkit for statistical relational learning, and show its use in a univariate case study.

About Sofus Macskassy: Dr. Sofus Attila Macskassy is a Senior Research Scientist at Fetch Technologies, Inc.

Last updated: Mon Jun 19 17:44:06 2006