A PATE-based Approach for Training Graph Neural Networks under Label Differential Privacy | IEEE Conference Publication | IEEE Xplore