VC dimension of partially quantized neural networks in the overparametrized regime.
Yutong WangClayton ScottPublished in: ICLR (2022)
Keyphrases
- vc dimension
- neural network
- upper bound
- sample complexity
- covering numbers
- concept classes
- lower bound
- sample size
- vapnik chervonenkis dimension
- statistical learning theory
- distribution free
- inductive inference
- generalization bounds
- back propagation
- empirical risk minimization
- artificial neural networks
- concept class
- pac learning
- pac learnability
- compression scheme
- uniform convergence
- euclidean space
- learning theory
- function classes
- compression algorithm
- model selection
- image quality
- np hard
- special case
- computational complexity
- image processing