Fig. 9
From: Regularized Simple Graph Convolution (SGC) for improved interpretability of large datasets
The distribution of the parameter values inferred for the Cora dataset with the application of the SGC and the regularized SGC. In a the SGC is applied and the histograms of the parameter values for each class in \(\varvec{\Theta }\) is shown in the plots. b Shows the equivalent plots but using the regularized SGC that penalizes the number of features. The majority of the features are around value zero