Speaker
Ambrus Tamás
(ELTE)
Description
Kernel methods are widely used in nonparametric statistics and machine learning. In this talk kernel mean embeddings of distributions will be used for the purpose of uncertainty quantification. The main idea of this framework is to embed distributions in a reproducing kernel Hilbert space, where the Hilbertian structure allows us to compare and manipulate the represented probability measures. We review some of the existing theoretical results and present new applications of this powerful tool. Distribution‐free, nonparametric results will be introduced for supervised learning problems (classification and regression).
| Classification | Mainly methodology |
|---|---|
| Keywords | Kernel methods, Statistical learning |
Primary author
Ambrus Tamás
(ELTE)