Events & News
Colloquium
| Category | Lecture |
| Date | Monday, November 10, 2008 |
| Time | 11:00 am |
| Location | GS 906 |
| Speaker | Dr. Xifeng Yan |
| Affiliation | IBM T. J. Watson Research Center |
Robust Methods for Mining Interesting Graph Patterns
*NOTE THAT THIS TALK IS ON A MONDAY*
Graphs become increasingly important in modeling complex structures that emerge in a wide range of domains. In the core of many graph-related applications, there is a strong need for mining interesting graph patterns with user-specified objective functions. Unfortunately, most interestingness functions are not anti-monotonic, for which existing frequency-centric mining algorithms may fail due to the exponential pattern space. In this talk, I will first introduce a new method that explores the correlation between structural similarity and interestingness similarity, which is able to identify the most significant patterns quickly by searching dissimilar patterns. The discovered patterns can be used to build high-quality interpretable graph classifiers. Next, I will present a new graph summarization technique that handles noise in large graphs and show its success in several areas including gene network analysis, software failure diagnosis, and social network analysis.
Biography
Xifeng Yan is a Research Staff Member at IBM T. J. Watson Research Center. He obtained his BE degree in computer engineering from Zhejiang University and his PhD degree in computer science from the University of Illinois at Urbana-Champaign. His current research interests are data mining, bioinformatics, and database systems. He has published more than 40 papers in refereed journals and conferences, including TODS, Bioinformatics, TSE, SIGMOD, SIGKDD, VLDB, and ISMB. His dissertation on graph mining and management received the 2006 ACM-SIGMOD Best Dissertation Runner-up Award.
www.xifengyan.net