Riemannian Laplace approximations for Bayesian neural networks
–Neural Information Processing Systems
However, practical posteriors are often, even locally, highly non-Gaussian, and empirical performance deteriorates.
Neural Information Processing Systems
Oct-8-2025, 19:21:56 GMT
- Country:
- Europe > Denmark (0.04)
- North America
- Canada > Ontario
- Toronto (0.14)
- United States
- California (0.04)
- Massachusetts > Suffolk County
- Boston (0.04)
- Canada > Ontario
- Technology: