Data are usually described and summarized by centers of mass. As an alternative, data points can be represented as linear combinations of extreme points of the data that lie on the convex hull of all data (aka the frame). These extreme points are however not always trivial to compute, particularly when the dimensionality of the space is high. In this talk, I will present exact and approximate techniques to compute the frame. The techniques are analyzed theoretically and empirically.
Ort: Seminarraum INB (Geb. 64, 1.OG, Raum 17)
