A Survey on Kernels for Structured Data

PhD Qualifying Examination


Title: "A Survey on Kernels for Structured Data"

Mr. Yang RUAN


Abstract:

In recent years, kernel methods have aroused increasing interest in the
machine learning community. The success of kernel methods is due to the
solid foundation built on mathematics and statistics as well as the good
performance when applied to real world problems. In a specific class of
problems that deal with structured data, the design of the kernel
functions, or kernels, plays an important role. For those problems,
kernels should be mathematically valid, computationally efficient and more
importantly, model the characteristics of data. In this survey, we review
several kernel construction methods for structured data, which is the
major kind of data from bioinformatics, chemistry and other applications.
Of the structured data, we focus on strings, graphs and hypergraphs.


Date:     		Friday, 7 December 2007

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

Venue:                  Room 3402
			lifts 17-18

Committee Members:      Dr. Dit-Yan Yeung (Supervisor)
			Dr. James Kwok (Chairperson)
			Dr. Brian Mak
			Dr. Nevin Zhang


**** ALL are Welcome ****