PAC-Bayes risk bounds for sample-compressed Gibbs classifiers.
François LavioletteMario MarchandPublished in: ICML (2005)
Keyphrases
- pac bayes
- risk bounds
- data dependent
- empirical risk minimization
- learning algorithm
- statistical learning theory
- vc dimension
- sample size
- stochastic processes
- markov random field
- generalization bounds
- support vector machine
- data compression
- machine learning
- linear classifiers
- active learning
- training data
- decision trees