A rationale from frequency perspective for grokking in training neural network.
Zhangchen ZhouYaoyu ZhangZhi-Qin John XuPublished in: CoRR (2024)
Keyphrases
- neural network
- training algorithm
- training process
- feed forward neural networks
- neural network training
- multi layer perceptron
- back propagation
- feedforward neural networks
- training patterns
- backpropagation algorithm
- recurrent networks
- test set
- low frequency
- artificial neural networks
- pattern recognition
- multi layer
- train a neural network
- neural network structure
- multilayer neural network
- self organizing maps
- genetic algorithm
- training set
- viewpoint
- supervised learning
- high frequency
- feed forward
- radial basis function network
- learning vector quantization
- radial basis function
- network model
- training samples
- neural network model
- backpropagation neural network
- neural network is trained
- error back propagation
- recurrent neural networks
- prediction model