【学术报告】Simultaneous Pattern and Data Clustering for Discrete-Valued Data

发稿时间:2007-03-23浏览次数:1932

                                           LAMDA 
                        http://lamda.nju.edu.cn

                Simultaneous Pattern and Data Clustering
                       for Discrete-Valued Data

 



报告人:Dr. Andrew K.C. Wong, IEEE Fellow
       University of Waterloo, Canada

时间:  3月26日(星期一),14:30-15:30

地点:  蒙民伟楼404会议室
 
摘要:

In data mining and knowledge discovery, pattern discovery is developed to discover previously unknown regularities in the data. However, the number of patterns produced by pattern discovery is usually overwhelming. To effectively manage the discovered patterns for further analysis from a large database, where interesting information and relevant patterns might be scattered, entangled and spanning in various data subspaces, is still a great challenge. This talk will present a new method to simultaneously cluster patterns and data to meet this challenge. It is different from conventional clustering in which the relation between the clustered data and the clustered patterns is established explicitly. This explicit data-pattern relation serves two purposes: data analysis and human interpretation. While data is grouped and broken down for further analysis, patterns are easy for human to interpret. The data-pattern relation enables users, on the one hand, to further analyze individual clusters and their relations via the data, and, on the other, to interpret why the data clusters are formed through the patterns they contain. Furthermore, within each cluster, noises are marginalized and irrelevant peripheral data are filtered out. Hence, the data can be analyzed more effectively. To achieve in-depth data analysis, the proposed method takes a divide-and-conquer approach. A large set of patterns is first clustered and each pattern cluster together with its associated data grouping is studied individually as supported by the explicit data-pattern relation. To evaluate the characteristics and the effectiveness of the proposed method, extensive experimental results on synthetic and real-world data are presented. Comparisons with other related methods are also given.

个人简历:

Dr. Andrew K.C. Wong currently is a Distinguished Professor Emeritus at the University of Waterloo where he is also an Adjunct Professor of the School of Computer Sciences and the Electrical and Computer Engineering Department.  He was the Founding Director of the renowned Pattern Analysis and Machine Intelligence Laboratory (PAMI Lab) and a Distinguished Chair Professor at the Hong Kong Polytechnic University (00-03). Dr. Wong holds a Ph.D. from Carnegie Mellon University; and a B.Sc (Hons) and M.Sc. from the Hong Kong University. He is an IEEE Fellow (for his contribution in machine intelligence, computer vision, and intelligent robotics).