Thursday
Friday
Organizers:
Regina Liu
Rutgers University
[email protected]
Robert Serfling
University of Texas at Dallas
[email protected]
Diane Souvaine
Tufts University
[email protected]
Yehuda Vardi
Rutgers University
[email protected]
Presented under the auspices of the DIMACS Special Focus on Data
Analysis and Mining and the DIMACS Special Focus on Computational
Geometry and Applications.
*********************************************************
Multivariate statistical methodology plays a role of ever increasing
importance in real life applications, which typically entail a host of
interrelated variables. Simple extensions of univariate statistics to
the multivariate setting do not properly capture the
higher-dimensional features of multivariate data, nor do they yield
geometric solutions because of the absence of a natural order for
multidimensional Euclidean space. A more promising approach is the one
based on "data depth", which can provide a center-outward ordering of
points in Euclidean space of any dimension. Extensive developments in
recent years have generated many attractive depth-based tools for
multivariate data analysis, with a wide range of applications. The
diversity in approaches, emphases, and concepts, however, makes it
necessary to seek unified views and perspectives that would guide the
further development of the depth-based approach.
The concept of data depth provides new perspectives to probabilistic
as well as computational geometries. In particular, the development of
implementable computing algorithms for depth-based statistics has
brought about many new challenges in computational geometry. This
workshop would create a unique environment for multidisciplinary
collaboration among computer scientists, theoretical and applied
statisticians, and data analysts. It would bring together active
researchers in these fields to discuss significant open issues,
establish perspective on applications, and set directions for further
research.