Risk bounds for the majority vote: from a PAC-Bayesian analysis to a learning algorithm.
Pascal GermainAlexandre LacasseFrançois LavioletteMario MarchandJean-Francis RoyPublished in: J. Mach. Learn. Res. (2015)
Keyphrases
- risk bounds
- bayesian analysis
- learning algorithm
- majority vote
- sample complexity
- vc dimension
- classifier ensemble
- computationally feasible
- classifier combination
- upper bound
- machine learning algorithms
- active learning
- fusion methods
- training data
- data dependent
- reinforcement learning
- machine learning
- random selection
- generalization ability
- statistical learning theory
- training samples
- generalization error
- classification algorithm
- training examples
- learning problems
- learning tasks
- supervised learning
- lower bound
- pattern recognition
- learning rate
- unlabeled data
- learning process