Group centrality and its applications

Abstract

The purpose of data mining is to efficiently identify important entities (e.g., data points, network nodes etc.) and discover hidden relationships between these entities from the data. Of course, the notion of importance as well as the interestingness of the discovered hidden relationships is highly application dependent. Therefore, in order to perform any data mining task effectively one needs to postulate the right question in order to identify relevant and interesting answers. In the first -- and largest -- part of the talk I will address the problem of identifying important nodes in networks and I will discuss two specific formulations of this general problem that are motivated by applications such as online learning platforms, social and information networks, the Web as well as computer networks and the Internet. In the second part of the talk I will discuss my ongoing research on discovering hidden relationships between network nodes, when the input consists of aggregate information that summarizes the hidden network structure as well as using combinatorial algorithms for detecting synonym words in text.

Bio

Dora Erdos is a PhD candidate in the department of Computer Science at Boston University. Her research interests lie in the area of algorithmic data mining. In 2008 she received her Bsc+Msc in pure mathematics from Eotvos University in Budapest. Between 2008-2009 she was a researcher at the Hungarian Academy of Sciences. During her PhD, Dora has been a visiting researcher at MPI for Informatics in Germany and a research intern at American Express in NYC.