A More Preliminary and Related Work A.1 More preliminary Graphon. A graphon [

Conventionally, we use a deterministic setting to select the node, e.g., the fixed grid However, these methods are unable to learn a new structure with a significantly reduced node size. We use the MSE [37] as the matching function Match() for simplicity. SGDD still be linear to the number of nodes in the original graph. We report the running time of the SGDD in the two datasets: Ogbn-arxiv, and Y elpChi. As shown in the Tab. 5, our approach achieves a similar running time to GCond when the We report the dataset statistics in Tab. 6. Citeseer-L: We use the Citeseer in the link prediction setting, named Citeseer-L.