DEGNN: Dual Experts Graph Neural Network Handling both Edge and Node Feature Noise.
Tai HasegawaSukwon YunXin LiuYin Jun PhuaTsuyoshi MurataPublished in: PAKDD (2) (2024)
Keyphrases
- neural network
- undirected graph
- edge weights
- graph structure
- directed graph
- weighted graph
- disjoint paths
- strongly connected
- edge orientation
- bipartite graph
- noise reduction
- artificial neural networks
- finding the shortest path
- densely connected
- random walk
- noise level
- nodes of a graph
- graph theory
- back propagation
- image features
- spanning tree
- structured data
- preprocessing stage
- learning algorithm
- complex networks
- fault diagnosis
- edge detection
- noise free
- vertex set
- image edges
- root node
- feature vectors
- gaussian noise
- approximation algorithms
- signal to noise ratio
- neural network model
- connected components
- train a neural network
- edge segments
- intensity variations
- graph representation
- edge information
- graph model
- domain experts
- missing data
- denoising
- genetic algorithm