Self-bounding Majority Vote Learning Algorithms by the Direct Minimization of a Tight PAC-Bayesian C-Bound.
Paul ViallardPascal GermainAmaury HabrardEmilie MorvantPublished in: ECML/PKDD (2) (2021)
Keyphrases
- majority vote
- pac bayesian
- upper bound
- rademacher complexity
- learning algorithm
- distribution free
- lower bound
- risk bounds
- generalization error
- generalization bounds
- vc dimension
- data dependent
- worst case
- sample complexity
- generalization ability
- classifier ensemble
- classifier combination
- learning problems
- fusion methods
- error bounds
- objective function
- active learning
- machine learning
- random selection
- concept class
- machine learning algorithms
- learning process
- learning rate
- statistical learning theory
- semi supervised learning
- labeled data
- learning tasks
- sample size
- unlabeled data
- model selection
- supervised learning
- training data