Piecewise-Linear Activations or Analytic Activation Functions: Which Produce More Expressive Neural Networks?
Anastasis KratsiosBehnoosh ZamanlooyPublished in: CoRR (2022)
Keyphrases
- activation function
- piecewise linear
- neural network
- feed forward neural networks
- feed forward
- hidden layer
- back propagation
- artificial neural networks
- neural nets
- dynamic programming
- multilayer perceptron
- learning rate
- radial basis function
- neural architecture
- network architecture
- multi layer perceptron
- machine learning
- basis functions
- fuzzy neural network
- fuzzy logic
- hyperplane
- training phase
- learning tasks
- recurrent neural networks
- input space
- fault diagnosis
- linear combination
- nearest neighbor
- pattern recognition