Pay Attention to Multi-Channel for Improving Graph Neural Networks

ChungYi Lin, Shen-Lung Tung, Winston H. Hsu

Keywords: Multi-channel Graph Attention, Spatial-Temporal Prediction, Graph Neural Networks

TL;DR: We propose Multi-channel Graph Attention (MGAT) to efficiently handle channel-specific representations, and verify its effectiveness by integrating it with various spatial-temporal graph neural networks to improve predictions.

Abstract: We propose Multi-channel Graph Attention (MGAT) to efficiently handle channel-specific representations encoded by convolutional kernels, enhancing the incorporation of attention with graph convolutional network (GCN)-based architectures. Our experiments demonstrate the effectiveness of integrating our proposed MGAT with various spatial-temporal GCN models for improving prediction performance.