Fig. 7
From: Regularized Simple Graph Convolution (SGC) for improved interpretability of large datasets
The results of applying the regularized SGC with an orthogonal constraint upon the projection achieved by using a change in regularization term \(L_3\). In a the vector projections within the methodology can be fit so that there is no loss and the orthogonality constraint is satisfied. In b the transformation upon the data with the network information is presented and how within this space the linear projections can separate the classes with the orthogonal vectors