PAC-Bayesian Theorems for Domain Adaptation with Specialization to Linear Classifiers.
Pascal GermainAmaury HabrardFrançois LavioletteEmilie MorvantPublished in: CoRR (2015)
Keyphrases
- domain adaptation
- linear classifiers
- generalization bounds
- multi class
- multiple sources
- cross domain
- hyperplane
- labeled data
- generalization error
- semi supervised learning
- semi supervised
- test data
- sentiment classification
- training data
- transfer learning
- svm classifier
- principal components
- unlabeled data
- k nearest neighbor
- prior knowledge
- neural network
- text classification
- hyperparameters
- supervised learning
- target domain
- co training
- high dimensional
- training set
- pattern recognition
- similarity measure