Fractional moment-preserving initialization schemes for training deep neural networks.
Mert GürbüzbalabanYuanhan HuPublished in: AISTATS (2021)
Keyphrases
- neural network
- training process
- training algorithm
- feedforward neural networks
- multi layer perceptron
- feed forward neural networks
- pattern recognition
- fuzzy logic
- error back propagation
- artificial neural networks
- back propagation
- lower order
- deep architectures
- radial basis function network
- neural network model
- backpropagation algorithm
- training set
- training phase
- multilayer neural network
- recurrent neural networks
- neural network structure
- multi layer
- feed forward
- training samples
- genetic algorithm
- machine learning
- network architecture
- hidden layer
- multilayer perceptron
- test set
- training examples
- active learning
- training data