Asymptotic Bayesian generalization error when training and test distributions are different.
Keisuke YamazakiMotoaki KawanabeSumio WatanabeMasashi SugiyamaKlaus-Robert MüllerPublished in: ICML (2007)
Keyphrases
- generalization error
- training set
- training error
- training and test sets
- supervised learning
- cross validation
- learning machines
- training set size
- model selection
- active learning
- learning algorithm
- upper bound
- binary classification
- classification error
- sample complexity
- training data
- test data
- linear classifiers
- test set
- sample size
- probability distribution
- worst case
- training examples
- risk minimization
- bayesian networks
- test cases
- subspace information criterion
- target function
- uniform convergence
- error rate
- maximum likelihood
- decision trees
- training process
- boosting algorithms
- training samples
- class distribution
- random variables
- naive bayes classifier
- large deviations
- support vector machine
- artificial neural networks
- conditional expectation
- generalization error bounds
- e learning