How Graph Neural Networks Learn: Lessons from Training Dynamics in Function Space.
Chenxiao YangQitian WuDavid WipfRuoyu SunJunchi YanPublished in: CoRR (2023)
Keyphrases
- neural network
- recurrent networks
- training process
- training algorithm
- function approximators
- pattern recognition
- recurrent neural networks
- feed forward neural networks
- equivalence classes
- multi layer perceptron
- feedforward neural networks
- space time
- feed forward
- neural network training
- function approximation
- graph theory
- graph representation
- learning rules
- directed graph
- lessons learned
- training examples
- backpropagation algorithm
- high dimensional
- training set
- nearest neighbor graph
- random walk
- low dimensional
- multi layer
- dynamic model
- entry level
- error back propagation
- directed acyclic graph
- bipartite graph
- graph matching
- neural network model
- self organizing maps
- structured data
- training samples
- supervised learning
- artificial neural networks
- training data
- genetic algorithm