Speaker:
Brent Davis, CSU Math
Title:
The MLV line for Pattern Recognition in High-Dimensional Data Sets
Abstract:
This talk will introduce an object to analyze a cluster of points embedded
on a union of Grassmann manifolds, called the MLV line of best fit. One
feature of the MLV line is the ability to capture common features of the
points. The MLV line is a solution to an optimization problem involving a
function of the principal angles between points. Two computational methods
will be discussed including homotopy methods. We will apply this method
to imaging data collected in the Pattern Analysis Lab. This is joint work with
Dan Bates, Michael Kirby, Justin Marks, and Chris Peterson.