Sparse Sobolev GNN (S2-GNN), employs Hadamard products between matrices to maintain the sparsity level in graph representations. S2-GNN utilizes a cascade of filters with increasing Hadamard powers to generate a diverse set of functions.
We evaluate the performance of S2-GNN on multiple datasets, which include tissue phenotyping in colon cancer histology images, text classification of news, activity recognition using sensors , recognition of spoken letters, as well as the commonly used node classification benchmarks Cora , Citeseer, Pubmed, and OGBN-proteins .
These diverse datasets provide a comprehensive evaluation of S2-GNN’s performance across different domains.