Eugene's ARGUS work (2004-2006)

The ARGUS project was a joint research project involving the Carnegie Mellon University and DYNAMiX Technologies. Jaime Carbonell at Carnegie Mellon was a principal investigator for this project, and Phil Hayes at DYNAMiX Technologies was a co-principal investigator. The purpose was to develop techniques for identification of both known and surprising patterns in large-scale databases, and apply these techniques to homeland security challenges. This work involved three main directions: I was working on the problem of identifying approximate matches to known patterns in large-scale databases, which involved indexing of large collections of structured records, representation of known patterns as database queries, and retrieval of approximate matches to given queries.

Publications

Eugene Fink, Aaron Goldstein, Philip Hayes, and Jaime G. Carbonell. Search for approximate matches in large databases. In Proceedings of the IEEE International Conference on Systems, Man, and Cybernetics, 2004. See PostScript, PDF, abstract, or conference talk.

B. Cenk Gazen, Jaime G. Carbonell, Philip J. Hayes, Chun Jin, and Eugene Fink. Hypothesis formation and tracking in ARGUS. NIMD Principal Investigator Meeting, 2004. See PostScript, PDF, or conference talk.

Jaime G. Carbonell, Eugene Fink, Chun Jin, B. Cenk Gazen, Santosh Ananthraman, Philip J. Hayes, Ganesh Mani, and Dwight Dietrich. Exploring massive structured data in ARGUS. NIMD Principal Investigator Meeting, 2005. See PostScript, PDF, or conference talk.

Jaime G. Carbonell, Eugene Fink, Chun Jin, B. Cenk Gazen, Johny Mathew, Abhay Saxena, Vini Satish, Santosh Ananthraman, Dwight Dietrich, and Ganesh Mani. Scalable data exploration and novelty detection. NIMD Principal Investigator Meeting, 2006. See PostScript, PDF, or conference talk.

Talks

ARGUS: Novelty detection and profile tracking from massive data (30 minutes). Presentation at the Sixth SIAM International Conference on Data Mining, April 21, 2006.


Back to my research, home page, or index.