Breaking the Activation Function Bottleneck through Adaptive Parameterization.
Sebastian FlennerhagHujun YinJohn A. KeaneMark ElliotPublished in: CoRR (2018)
Keyphrases
- activation function
- neural network
- hidden layer
- back propagation
- artificial neural networks
- network size
- neural nets
- feed forward
- feed forward neural networks
- sigmoid function
- learning rate
- multilayer perceptron
- basis functions
- feedforward neural networks
- hidden nodes
- training phase
- small number
- extreme learning machine
- chaotic neural network