A Strongly Quasiconvex PAC-Bayesian Bound.
Niklas ThiemannChristian IgelOlivier WintenbergerYevgeny SeldinPublished in: ALT (2017)
Keyphrases
- pac bayesian
- quasiconvex
- distribution free
- rademacher complexity
- multicriteria optimization
- finite number
- objective function
- data dependent
- generalization bounds
- utility function
- error bounds
- convex functions
- convex sets
- large deviations
- lower bound
- risk bounds
- learning theory
- model selection
- upper bound
- normal distribution
- vc dimension
- ranking algorithm
- worst case
- training set