Wednesday, April 17, 2013
5:00 PM
Hubbel Auditorium
Mauro Maggioni
Duke University
Geometry and Analysis of High Dimensional Data
I will introduce several high-dimensional data problems
that arise in a variety of disciplines, from machine learning
and computer vision to computational chemistry and
network analysis. I will then discuss a common mathematical
language and ideas to attack such problems—based on diffusion
geometry, multiscale geometric analysis, and graph theory—that yield practical efficient algorithms for sifting through large data sets and learn properties of interest. Much of this mathematical language is at the intersection of analysis, probability, and geometry. I will emphasize the connections between apparently different problems and some of the key ideas in the geometric analysis of high- dimensional data that provide efficient ways of attacking these problems.
This talk is directed to a wide audience of scientists.