XAI for Graphs: Explaining Graph Neural Network Predictions by Identifying Relevant Walks.
Thomas SchnakeOliver EberleJonas LedererShinichi NakajimaKristof T. SchüttKlaus-Robert MüllerGrégoire MontavonPublished in: CoRR (2020)
Keyphrases
- neural network
- graph theory
- graph representation
- graph construction
- graph structure
- labeled graphs
- graph databases
- directed graph
- graph clustering
- graph theoretic
- weighted graph
- graph mining
- graph matching
- graph model
- graph structures
- graph classification
- graph theoretical
- subgraph isomorphism
- graph search
- undirected graph
- adjacency matrix
- artificial neural networks
- graph properties
- graph partitioning
- graph data
- bipartite graph
- social graphs
- connected graphs
- random graphs
- structural pattern recognition
- spanning tree
- series parallel
- back propagation
- graph isomorphism
- strongly connected
- disk resident
- minimum spanning tree
- graph transformation
- graph kernels
- web graph
- graph representations
- connected dominating set
- reachability queries
- query graph
- attributed graphs
- graph embedding
- dynamic graph
- random walk
- proximity graph
- edge weights
- graph layout
- bounded treewidth
- graph patterns
- maximum cardinality
- maximum independent set
- dense subgraphs
- maximum common subgraph
- massive graphs
- graph drawing
- social networks
- polynomial time complexity
- average degree
- maximum clique
- association graph
- topological information
- planar graphs
- connected components
- neural network model
- fully connected
- community detection
- knn
- densely connected
- inexact graph matching
- evolving graphs
- quasi cliques
- frequent subgraphs
- maximal cliques