home · mobile · calendar · bactac · 2006-2007 · 

BACTAC - Bijral

Generative Models for Spectral Clustering on a Hypershere
Avleen Bijral Jr.
Master's Candidate

Many machine learning applications involve data, that can be analyzed as unit vectors on a d-dimensional Hypersphere, or equivalently are directional in nature. Spectral clustering techniques generate embeddings that constitute an example of directional data and can result in different topologies on a Hypersphere (depending on the original topology). Other examples of directional data includes text and some sub-domains of bioinformatics. For example the Watson and von Mises Fisher distributions for directional data presents a tractable form. In this talk I will be talking about generative models for these distributions and an EM algorithm that generates posterior cluster probabilities for points.

Department of Computer Science
University of Colorado Boulder
Boulder, CO 80309-0430 USA
May 5, 2012 (14:24)