Steve Oudot: Topological features for data
Tid: Må 2019-02-04 kl 10.00 - 12.00
Föreläsare: Steve Oudot, Inria
Plats: Room 3418, math department KTH, floor 4 (bottom floor)
Abstract: One of the main selling points of Topological Data Analysis (TDA) is to provide new families of features for data, able to capture structural information of a different kind, with provable stability and invariance guarantees. In this lecture I will review some of the mainstream approaches to design topological descriptors, and to turn these descriptors into feature vectors. While the exposition will
concentrate mostly on the theoretical guarantees associated with these features, I will also mention in passing some of their real-life applications, mostly in supervised learning tasks.