When

Start: 01/27/2016 - 4:15pm

End : 01/27/2016 - 4:15pm

End : 01/27/2016 - 4:15pm

Category

Colloquium

Speaker

Andrea Bertozzi (UCLA)

Abstract

This talk is about current research in algorithms for "big data". We will focus on two classes of models - (1) topic modeling for document classification (2) algorithms based on graphical models used for hard clustering of big data. Many of these methods can be understood with a basic undergraduate math background including linear algebra. For topic modeling we will discuss basic models like nonnegative matrix factorication and latent dirichlet allocation. For the graphical methods we introduce the graph Laplacian and present spectral clustering methods and also some recent graph-cut based methods developed at UCLA. Examples of data are diverse from hyperspectral imagery to twitter data

Where

Argue Auditorium, Millikan, Pomona College