Polynomial bounds for VC dimension of sigmoidal neural networks.
Marek KarpinskiAngus MacintyrePublished in: STOC (1995)
Keyphrases
- vc dimension
- neural network
- vapnik chervonenkis dimension
- upper bound
- vapnik chervonenkis
- covering numbers
- lower bound
- sample complexity
- sample size
- concept classes
- generalization bounds
- statistical learning theory
- distribution free
- worst case
- compression scheme
- inductive inference
- empirical risk minimization
- concept class
- pac learning
- learning machines
- artificial neural networks
- function classes
- back propagation
- uniform convergence
- risk bounds
- euclidean space
- dnf formulas
- target function
- mistake bound
- image compression
- support vector machine
- optimal solution