Formal limitations of sample-wise information-theoretic generalization bounds.
Hrayr HarutyunyanGreg Ver SteegAram GalstyanPublished in: ITW (2022)
Keyphrases
- information theoretic
- generalization bounds
- data dependent
- mutual information
- information theory
- learning theory
- generalization ability
- model selection
- vc dimension
- ranking algorithm
- sample size
- statistical learning theory
- learning problems
- linear classifiers
- ranking functions
- learning algorithm
- kernel machines
- training data
- neural network
- data mining
- machine learning algorithms
- supervised learning
- multi class
- lower bound
- feature selection
- machine learning