Seminar: Mining and Modeling in Networks from Diverse Domains

Dr. Petko Bogdanov (University of California, Santa Barbara)

Monday, March 3, 2014, 10:30AM-12PM, EV 3.309

Abstract

Graphs can represent relationships between data entities and hence provide an expressive model for big data produced by real-world systems. A number of big data domains feature an inherent network structure that can be modeled as a graph - social networks and media, transportation networks, gene networks, the brain, communication and information networks. Studying, understanding and predicting the inherent processes in all those areas requires scalable graph mining algorithms and models. While a well-suited representation model, graphs present unique algorithmic challenges. Real-world networks feature more than just the structure among entities - they evolve over time and incorporate content and features associated with nodes and edges. To address those challenges, we develop novel formulations and algorithms that scale to large instances while not compromising the quality of mined results.

In my talk, I will present my research on subgraph mining in time-evolving networks and modeling and predicting user behavior in social media. I will also discuss future research directions in this area.

Bio

Dr. Petko Bogdanov is a postdoctoral fellow in the Computer Science department at University of California, Santa Barbara. He is also affiliated with the Network Science Collaborative Technology Alliance (NS-CTA). His research interests are in scalable data mining, data modeling and data management with a focus on graph data and with interdisciplinary applications in bioinformatics, sociology, neuroscience and materials research. He received his PhD and MS in Computer Science from UC Santa Barbara.