Emerging Trends in Graph WaveNet Research
Seungyeop Song, Kihyun Seol, Y.H. Lee, Heejae Park, Laihyuk Park · 2024
Graph Neural Networks (GNNs) have been developed to learn the spatial and temporal patterns inherent in graph-structured data, patterns that are challenging to model with traditional machine learning methods. GNNs excel at capturing complex relationships and dependencies between nodes in a graph. However, most existing GNN-based methods depend on a fixed graph structure to capture spatial dependencies. To address this limitation, Graph WaveNet was proposed, introducing a self-adaptive adjacency matrix to overcome the constraints of fixed graph structures. In this paper, we delve into the architecture of Graph WaveNet and examine its emerging research trends.