Bayesian co-clustering and its application

Speaker: Dr. Tomoki Tokuda(Researcher, OIST )

Date: 30th Oct.(Th.)

Time:10:30~11:30

Venue:C016(Lab1)

Title:Bayesian co-clustering and its application

Abstract:

In this talk, I discuss a novel approach for the dimension reduction of high dimensional data. The method is based on nonparametric Gaussian co-clustering, in which we assume that in each cluster block, the instances follow an i.i.d. univariate Gaussian distribution. It will be shown that this model can fit a specific class of multivariate Gaussian distributions with exchangeable features. Next, we consider application of this model to defining functional connectivity between brain regions in analyzing fMRI (functional Magnetic Resonance Imaging) data. Both advantage and disadvantage of this application will be discussed