Reducing Oversmoothing in Graph Neural Networks by Changing the Activation Function.
Dimitrios KelesisDimitrios VogiatzisGeorgios KatsimprasDimitris FotakisGeorgios PaliourasPublished in: ECAI (2023)
Keyphrases
- activation function
- neural network
- back propagation
- artificial neural networks
- feed forward
- feed forward neural networks
- hidden layer
- neural nets
- multilayer perceptron
- hidden neurons
- backpropagation algorithm
- learning rate
- feedforward neural networks
- basis functions
- extreme learning machine
- hidden nodes
- single layer
- sigmoid function
- radial basis function
- fuzzy neural network
- network architecture
- single hidden layer
- hidden units
- training phase
- training algorithm
- rbf neural network
- graphical models
- fuzzy logic
- probabilistic model