Non-asymptotic approximations of Gaussian neural networks via second-order Poincaré inequalities.
Alberto BordinoStefano FavaroSandra FortiniPublished in: CoRR (2023)
Keyphrases
- neural network
- pattern recognition
- higher order
- neural network model
- artificial neural networks
- linear inequalities
- worst case
- difference equations
- multi layer
- feed forward
- closed form
- fault diagnosis
- maximum likelihood
- closed form expressions
- akaike information criterion
- expected values
- central limit theorem
- differential operators
- back propagation
- fuzzy logic
- gaussian mixture
- neural nets
- planar graphs
- expectation propagation
- self organizing maps
- finite sets
- sufficient conditions
- multiscale