Maximal Discrepancy vs. Rademacher Complexity for error estimation.
Davide AnguitaAlessandro GhioLuca OnetoSandro RidellaPublished in: ESANN (2011)
Keyphrases
- error estimation
- rademacher complexity
- generalization error
- data dependent
- model selection
- cross validation
- error estimates
- upper bound
- active learning
- training set
- learning algorithm
- training data
- error bounds
- risk bounds
- sample complexity
- supervised learning
- sample size
- kernel methods
- support vector machine
- risk minimization
- manifold regularization
- text classification
- feature vectors
- reinforcement learning
- data mining