Fully Distributed Online Training of Graph Neural Networks in Networked Systems
This enables more adaptable and faster development of decentralized AI for large-scale networked systems like wireless networks and power grids, representing a novel method rather than an incremental improvement.
The paper tackles the limitation of centralized training for graph neural networks (GNNs) in networked systems by developing a fully distributed online training approach, demonstrating its effectiveness in tasks like node regression and wireless network optimization with minimal communication overhead.
Graph neural networks (GNNs) are powerful tools for developing scalable, decentralized artificial intelligence in large-scale networked systems, such as wireless networks, power grids, and transportation networks. Currently, GNNs in networked systems mostly follow a paradigm of `centralized training, distributed execution', which limits their adaptability and slows down their development cycles. In this work, we fill this gap for the first time by developing a communication-efficient, fully distributed online training approach for GNNs applied to large networked systems. For a mini-batch with $B$ samples, our approach of training an $L$-layer GNN only adds $L$ rounds of message passing to the $LB$ rounds required by GNN inference, with doubled message sizes. Through numerical experiments in graph-based node regression, power allocation, and link scheduling in wireless networks, we demonstrate the effectiveness of our approach in training GNNs under supervised, unsupervised, and reinforcement learning paradigms.