RapidGNN: Energy and Communication-Efficient Distributed Training on Large-Scale Graph Neural Networks