Christos Faloutsos

Christos Faloutsos


Office 7003 Gates and Hillman Centers


Phone (412) 268-1457

Computer Science Department


Administrative Support Person
Oliver Moss

Research Statement

There are two main focus areas: graph mining and stream mining. In the first, the goal is to find patterns in large graphs, so that we can spot anomalies, communities, patterns and regularities. Graphs appear in many instances: as document-term bipartitegraphs in Information retrieval, as web pages or blogs linking to each other, as customer-product recommendations, as protein-protein regulatory networks, as computer-network traffic, and many more. Our emphasis is on scalability, so that we can handle graphs withthousands and millions of nodes. Research directions include time-evolving graphs, where we have beenusing 'tensors' to find patterns, as well as graphs where the nodes and/or the edges have attributes.

The second research area focuses on streams, which are semi-infinitenumerical time series. The setting also has numerous applications, like sensor data monitoring, motion capture data, automatic alerts in the 'self-*' PetaByte storage system, chlorine level monitoring on the drinking water, and several more. The emphasis is to develop algorithms that inspect every measurementonly once, and then discard it, since we can not affort to store the huge volume of historical data.

The common threads in both areas are the power-laws and the existenceof self-similarity. Real graphs have skewed, Zipf-like degree distributions, and consist of communities-within-communities. Similarly, real sensor measurements are often bursty, but still self-similar, with bursts within bursts. We use or develop tools that exactly exploit the power laws and self-similarity, to find better patterns and anomalies than standard tools would find.

keywords: Database Management Systems, Data Mining, Graphs, Social Networks, Network Security.

Recent Publications

Christos Faloutsos ( 2022 ) Knowledge and Information Systems, Autonomous graph mining algorithm search with best performance trade-off , Vol: 64 Issue: ( 6 ) , Page(s): 1571- 1602 .

Park N, Rossi R, Koh E, Burhanuddin IA, Kim S, Du F, Ahmed N, Faloutsos C ( 2022 ) WWW 2022 - Proceedings of the ACM Web Conference 2022, CGC: Contrastive Graph Clustering forCommunity Detection and Tracking , Page(s): 1115 - 1126

Li J, Zhao T, Li J, Chan J, Faloutsos C, Karypis G, Pantel SM, McAuley J ( 2022 ) SIGIR 2022 - Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, Coarse-to-Fine Sparse Sequential Recommendation , Page(s): 2082 - 2086

Ito H, Faloutsos C ( 2022 ) Proceedings of the 2022 SIAM International Conference on Data Mining, SDM 2022, DualCast: Friendship-Preference Co-evolution Forecasting for Attributed Networks , Page(s): 46 - 54

Recent Awards

Pacific Asia Conference on Knowledge Discovery and Data Mining Most Distinguished Contribution (PAKDD) | 2018 | Test of Time Award
The Steering Committee considers it an honor to award the PAKDD Distinguished Contributions Award for 2018 to Professor Christos Faloutsos, for his many seminal contributions to the field of data mining, including time series matching, network analysis, graph computation, and their scalability. Especially notable is his highly successful program in showing how strong mathematical results can be used to design highly original and novel data mining algorithms. He has also provided tremendous service to raising the visibility of PAKDD conference through publishing multiple high impact papers over the years.