Accelerating Sample-based GNN Training by Feature Caching on GPUs | IEEE Conference Publication | IEEE Xplore