Self-Distillation Amplifies Regularization in Hilbert Space.
Hossein MobahiMehrdad FarajtabarPeter L. BartlettPublished in: CoRR (2020)
Keyphrases
- hilbert space
- reproducing kernel
- reproducing kernel hilbert space
- von neumann
- finite dimensional
- infinite dimensional
- kernel function
- euclidean space
- scale spaces
- multivariate time series
- continuous functions
- convex sets
- loss function
- learning theory
- data dependent
- kernel methods
- vector valued
- feature maps
- density estimation
- feature space
- expected utility
- multiscale
- higher dimensional