CS Colloquium on Monday, Mar 22 at 11AM

Title: Computational Geometry and Statistical Depth Measures

Speaker: Diane Souvaine
         Computer Science Department
         Tufts University
         http://www.cs.tufts.edu/~dls/

Place: MCS 135, 111 Cummington Street (please see
       http://cs-www.bu.edu/colloquium for directions)

-------------------------------------------------------------------------  

Abstract:

Most real life experiments are high-dimensional (multivariate) by
nature and large-scale multivariate datasets are now made tractable by
recent explosive advances in computer technology. Consequently 
good analytical tools are needed to process and understand these
increasingly large data sets.  Data depth is a statistical
analysis method that is based on the shape of the data. It has an
significant potential as an analysis method for real life data sets
because it does not require prior assumptions on the probability
distribution of data and deals with outliers.  Data-depth is 
inherently geometric.  Consequently, techniques from computational 
geometry, a subfield of algorithms which investigates geoometric problems,
can be used to compute data depth and associated metrics.  This talk will
describe some of the underlying geometric technigues, detail the positive 
results with respect to halfspace (Tukey) depth or regression depth, 
and present partial results and open problems related to simplicial depth.

-------------------------------------------------------------------------

Host: Stan Sclaroff (http://www.cs.bu.edu/~sclaroff)