Large Scale Kernel Based Clustering Algorithms

PhD Thesis Proposal Defence


Title: "Large Scale  Kernel Based Clustering Algorithms"

by

Mr. Kai Zhang


Abstract:

Clustering is an unsupervised data exploration scenario that is of
fundamental importance to pattern recognition and machine learning. This
proposal involves two types of clustering paradigms, the mixture model
clustering and graph-based clustering methods, with the primary focus on
how to improve the scaling behaviors of related algorithms for large scale
application. With regard to mixture models, we are interested in how to
reduce the model complexity in terms of number of components. We propose a
unified algorithm to simultaneously solve ``model simplification'' and
``component clustering'', and apply it with success in a number of
learning algorithms using mixture models, such as density based clustering
and and SVM testing. In the second part, we analyze the Nystrom method for
solving large eigenvalue problems. We identify one of its key inefficiency
and the propose a density weighted variant that greatly improves the
approximation quality. We test our approach on two important learning
problems, the normalized-cut and kernel principal component analysis and
obtain encouraging performance. In the future we shall further extend this
for more efficient and accurate low rank decomposition of kernel matrix.


Date:     		Friday, 15 February 2008

Time:                   2:00p.m.-4:00p.m.

Venue:                  Room 3588
			lifts 27-28

Committee Members:      Dr. James Kwok (Supervisor)
			Dr. Nevin Zhang (Chairperson)
                        Prof. Qiang Yang
			Dr. Dit-Yan Yeung


**** ALL are Welcome ****