SERF: Towards better training of deep neural networks using log-Softplus ERror activation Function.
Sayan NagMayukh BhattacharyyaPublished in: CoRR (2021)
Keyphrases
- activation function
- neural network
- feed forward neural networks
- backpropagation algorithm
- error function
- feedforward neural networks
- single hidden layer
- back propagation
- hidden layer
- training phase
- multi layer perceptron
- artificial neural networks
- feed forward
- training algorithm
- hidden neurons
- multilayer perceptron
- neural nets
- learning rate
- training process
- network architecture
- output layer
- radial basis function
- basis functions
- hidden nodes
- extreme learning machine
- fuzzy neural network
- recurrent neural networks
- training set
- pattern recognition
- single layer
- sigmoid function
- small number
- decision trees
- fuzzy logic
- learning algorithm