Input Feature Pruning for Accelerating GNN Inference on Heterogeneous Platforms | IEEE Conference Publication | IEEE Xplore