A research team has introduced a new out-of-core mechanism, Capsule, for large-scale GNN training, which can achieve up to a 12.02× improvement in runtime efficiency, while using only 22.24% of ...