Depth separation and weight-width trade-offs for sigmoidal neural networks.
Amit DeshpandeNavin GoyalSushrut KarmalkarPublished in: ICLR (Workshop) (2018)
Keyphrases
- neural network
- trade off
- activation function
- feed forward neural networks
- feedforward neural networks
- weight update
- back propagation
- artificial neural networks
- genetic algorithm
- neural network model
- pattern recognition
- hidden nodes
- training algorithm
- depth map
- fault diagnosis
- fuzzy systems
- depth information
- feed forward
- self organizing maps
- fuzzy logic
- lower bound
- rule extraction
- machine learning
- multi layer perceptron
- neural nets
- multi objective
- multilayer perceptron
- high quality
- extreme learning machine
- learning algorithm
- synaptic weights