Fig. 5
From: Regularized Simple Graph Convolution (SGC) for improved interpretability of large datasets
The results of applying the SGC method on the linearly inseparable data is presented here. a Plots the data points and the 2 learned parameter vectors by SGC methodology under different initializations. It can be seen how although the displayed classification vectors within \(\varvec{\Theta }\) with the network information provide the ability for a lossless prediction. Similarly, for b it can be seen how a new set of axes for the projection \({\mathbf {S}}^2 {\mathbf {X}}\) (network and features) the linear separation becomes visible