Hypothesis Transfer Learning with Surrogate Classification Losses: Generalization Bounds through Algorithmic Stability.
Anass AghbalouGuillaume StaermanPublished in: ICML (2023)
Keyphrases
- generalization bounds
- transfer learning
- algorithmic stability
- generalization ability
- machine learning
- data dependent
- model selection
- learning theory
- learning tasks
- learning algorithm
- generalization error
- learning problems
- text classification
- vc dimension
- machine learning algorithms
- uniform convergence
- ranking algorithm
- active learning
- statistical learning theory
- support vector machine
- semi supervised learning
- labeled data
- learning machines
- linear classifiers
- support vector
- kernel machines
- ranking functions
- training set
- support vector machine svm
- supervised classification
- multi task
- domain adaptation
- cross domain
- decision trees
- multi task learning
- feature selection
- sample complexity
- feature space
- target domain
- text categorization
- supervised learning
- ensemble learning
- hyperparameters
- upper bound
- ensemble methods
- feature vectors
- learning experience
- sample size
- class labels