We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
John Kent's contribution to the Discussion of the 'Discussion Meeting on Probabilistic and statistical aspects of machine learning'.
- Authors
Kent, John T
- Abstract
The document is a contribution to a discussion on probabilistic and statistical aspects of machine learning. The author focuses on diffusions and directional distributions in Euclidean space and compact Riemannian manifolds. They discuss the use of an Ornstein-Uhlenbeck process with specific parameters for the diffusion in Euclidean space and suggest alternative diffusions for manifolds. The author also mentions the von Mises-Fisher distribution and the Watson distribution as equilibrium distributions for specific drifts. They propose that the isotropic and anisotropic matrix Fisher distributions could be useful alternatives in certain settings.
- Subjects
MACHINE learning; EQUATIONS of motion; ORNSTEIN-Uhlenbeck process; RIEMANNIAN manifolds; BROWNIAN motion
- Publication
Journal of the Royal Statistical Society: Series B (Statistical Methodology), 2024, Vol 86, Issue 2, p324
- ISSN
1369-7412
- Publication type
Proceeding
- DOI
10.1093/jrsssb/qkad156