FR-GNN: Mitigating the Impact of Distribution Shift on Graph Neural Networks via Test-Time Feature Reconstruction.
Rui DingJielong YangFeng JiXionghu ZhongLinbo XiePublished in: IEEE Internet Things J. (2024)
Keyphrases
- neural network
- point distribution
- graph representation
- artificial neural networks
- random walk
- image features
- test data
- back propagation
- null hypothesis
- reconstruction process
- pattern recognition
- image reconstruction
- directed graph
- bipartite graph
- graph partitioning
- vector representation
- directed acyclic graph
- multi layer
- weighted graph
- statistical tests
- graph structure
- graph theory
- high resolution
- graph matching
- data distribution
- structured data
- test cases
- recurrent neural networks
- input pattern
- test statistic
- preprocessing stage
- statistical significance
- graph databases
- feature vectors
- three dimensional
- learning algorithm
- genetic algorithm