Computer Science Seminar Series
Ranking on Graph Data Using Kernelized Spatial Depth, with Application to Protein-Protein Interaction Data
3:00 p.m. Wednesday, October 8, 2008
Weir Hall, Room 235
Lan Gao
Ph.D. Student
Department of Mathematics
University of Mississippi
Abstract:
In many scientific fields, data is represented as a graph, where vertices corresponding to the objects and edges encode similarity among objects. Ranking and further clustering the vertices in a graph is very important conceptually and practically as many "real world" data can be regarded as graph, such as social networks, biology networks, and many kinds of computer networks including the World Wide Web. A vertex in these graph data may represent a person, a Protein, a web or any other objects. Graph ranking is about computing the rank of all the vertices in the graph which allows us to determine the most important vertices or cluster the vertices. In this research, we developed a new ranking algorithm which provides a real - valued ranking function on graph data by generating the kernel spatial depth (KSD) of each vertex. We applied KSD ranking method to the network data Protein-Protein data which combines various data sets from Database of Interaction Proteins. Moreover KSD ranking is compared with other ranking methods using the evaluate technique modified Kendall's Tau correlation coefficient.