Stronger generalization bounds for deep nets via a compression approach.
Sanjeev AroraRong GeBehnam NeyshaburYi ZhangPublished in: CoRR (2018)
Keyphrases
- generalization bounds
- data dependent
- learning theory
- generalization ability
- convex combinations
- model selection
- vc dimension
- linear classifiers
- compression scheme
- statistical learning theory
- learning problems
- data compression
- compression algorithm
- ranking algorithm
- image compression
- compression ratio
- ranking functions
- sample size
- neural network
- artificial neural networks
- support vector machine
- lower bound
- pattern recognition