Orthogonal-Padé Activation Functions: Trainable Activation functions for smooth and faster convergence in deep networks.
Koushik BiswasShilpak BanerjeeAshish Kumar PandeyPublished in: CoRR (2021)
Keyphrases
- activation function
- faster convergence
- network size
- neural network
- learning rate
- artificial neural networks
- feed forward
- hidden layer
- convergence speed
- neural nets
- back propagation
- neural architecture
- global optimum
- step size
- multilayer perceptron
- pso algorithm
- convergence rate
- radial basis function
- basis functions
- network architecture
- global optimization
- rbf neural network
- particle swarm optimization
- bp neural network
- training phase
- probabilistic model
- expert systems
- image processing