A PAC-Bayesian Generalization Bound for Equivariant Networks.
Arash BehboodiGabriele CesaTaco CohenPublished in: CoRR (2022)
Keyphrases
- pac bayesian
- generalization bounds
- distribution free
- data dependent
- rademacher complexity
- learning theory
- generalization ability
- vc dimension
- model selection
- error bounds
- large deviations
- learning problems
- ranking algorithm
- upper bound
- statistical learning theory
- learning machines
- uniform convergence
- linear classifiers
- concept classes
- normal distribution
- empirical risk minimization
- training set