1 paper · 1 filter
Jiaqi Wang, Xinyi Wu, James Cheng +1
Deep graph neural networks (GNNs) often suffer from oversmoothing, where node representations become overly homogeneous with increasing depth. While techniques like normalization,…