Bayesian Quadrature on Riemannian Data Manifolds

Christian Frohlich 1 Alexandra Gessner * 1 2 Philipp Hennig 1 2 Bernhard Scholkopf 2 Georgios Arvanitidis * 2

             Abstract
   Riemannian manifolds provide a principled way
   to model nonlinear geometric structure inherent
   in data. A Riemannian metric on said manifolds
   determines geometry-aware shortest paths and
   provides the means to define statistical models ac-
   cordingly. However, these operations are typically
   comp ...