Christos Faloutsos

Fredkin University Professor of Computer Science

Website

CMU Scholars Page

ORCiD

Office 7003 Gates and Hillman Centers

Email christos@andrew.cmu.edu

Phone (412) 268-1457

Department
Computer Science Department

Administrative Support Person
Oliver Moss

Research Interests
Artificial Intelligence
Machine Learning
Systems
Databases
Distributed Systems

Advisees
Jeremy Lee
Catalina Vajiac
Saranya Vijayakumar

CSD Courses Taught

15826 - Fall, 2025

15826 - Fall, 2024

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.

Publications