Integration of GAT and GCN for WiFi Positioning

초록

In this paper, we propose a novel algorithm that combines graph attention network (GAT) with graph convolutional network (GCN), both of which are types of graph neural networks (GNNs). GNNs have gained attention for their ability to overcome the non-line-of-sight (NLoS) bias in WiFi positioning by efficiently extracting local features from graphs. Existing GNN-based algorithms often rely on GCNs due to their low computational complexity. Our empirical experiments demonstrate that the proposed technique reduces RMSE by 8.6% and 21.0% for self-supervised and semi-supervised learning, respectively, compared to using GCN alone. © 2024 IEEE.

키워드