Characterising Structural Variations in Graphs
|Venue:||TextGraphs-5 Workshop at ACL 2010 in Uppsala|
|Date:||July 16, 2010|
pdf presentation slides (5.7 MB)
flv flash video (240 MB, 368 * 208 pixels)
mpeg2 video (720 MB, 640 * 360 pixels)
The talk commences by discussing some of the problems that arise when machine learning is applied to graph structures. A taxonomy of different methods organised around a) clustering b) characterisation and c) constructing generative models in the graph domain is introduced. With this taxonomy in hand, Dr. Hancock then describes a number of graph-spectral algorithms that can be applied to solve the many different problems inherent to graphs, drawing examples from computer vision research.