Geodesic Distance Between Graphs: A Spectral Metric for Assessing the Stability of Graph Neural Networks.
Soumen Sikder ShuvoAli AghdaeiZhuo FengPublished in: CoRR (2024)
Keyphrases
- geodesic distance
- neighborhood graph
- edge weights
- heat kernel
- neural network
- shortest path
- weighted graph
- spectral decomposition
- euclidean space
- euclidean distance
- graph construction
- fisher information
- maximum common subgraph
- manifold learning
- graph structure
- distance transform
- adjacency matrix
- graph mining
- geometric structure
- distance metric
- riemannian manifolds
- pattern recognition
- graph partitioning
- graph laplacian
- graph kernels
- spanning tree
- dimensionality reduction
- knn
- computer vision