Higher-Order Explanations of Graph Neural Networks via Relevant Walks.
Thomas SchnakeOliver EberleJonas LedererShinichi NakajimaKristof T. SchüttKlaus-Robert MüllerGrégoire MontavonPublished in: IEEE Trans. Pattern Anal. Mach. Intell. (2022)
Keyphrases
- higher order
- neural network
- high order
- genetic algorithm
- higher order relations
- pattern recognition
- graph structure
- artificial neural networks
- graph representation
- random walk
- pairwise
- natural images
- markov random field
- fault diagnosis
- directed graph
- neural network model
- bipartite graph
- weighted graph
- graph partitioning
- graph model
- multi layer
- social networks
- neural nets
- recurrent neural networks
- multilayer perceptron
- fuzzy logic
- graph theory
- link analysis
- directed acyclic graph
- conditional random fields
- back propagation
- graph theoretic
- low order
- graph based algorithm