The Connection Between Approximation, Depth Separation and Learnability in Neural Networks.
Eran MalachGilad YehudaiShai Shalev-ShwartzOhad ShamirPublished in: CoRR (2021)
Keyphrases
- neural network
- artificial neural networks
- pattern recognition
- approximation algorithms
- fuzzy logic
- back propagation
- depth map
- learning algorithm
- depth information
- boolean functions
- uniform distribution
- activation function
- finite automata
- approximation error
- multi layer
- approximation methods
- associative memory
- agnostic learning
- inductive logic programming
- neural network model
- error bounds
- radial basis function
- recurrent neural networks
- multilayer perceptron
- fuzzy systems
- training algorithm
- closed form
- self organizing maps
- sufficient conditions
- statistical queries
- reinforcement learning