Visual cluster analysis
Chris Leckie (The University of Melbourne)
NICTA SML SEMINARDATE: 2013-08-08
TIME: 11:15:00 - 12:15:00
LOCATION: NICTA - 7 London Circuit
CONTACT: JavaScript must be enabled to display this email address.
ABSTRACT:
While clustering techniques are immensely popular for tasks such as market segmentation studies, genomics and medical imaging, an ongoing challenge is how to interpret the set of clusters that have been found on a given data set. Moreover, before we start the computationally expensive task of clustering, can we easily ascertain any properties of the data set that might help focus our search for clusters, such as the likely number of clusters? Visual approaches to cluster analysis aim to help with these challenges. In this talk, I will focus on one class of approaches that use visualisations of reordered dissimilarity matrices. Iall briefly overview the history of these techniques, illustrate some of the optimisation problems underlying these techniques, and discuss some of our recent work on visual co-clustering and tracking cluster evolution.
BIO:
http://ww2.cs.mu.oz.au/~caleckie/
