Fig. 2
From: Regularized Simple Graph Convolution (SGC) for improved interpretability of large datasets
15 data points of 2 classes are randomly generated and form 2 distinct clusters residing on the opposite sides of the axis. a Presents the scatter plots of these data points. b Shows the network connections among those data points. In terms of a distance metric between the points, or from using a linear projection, erroneous label assignments can arise from using only a data but the incorporation of the network associations shown in b allows this to be bypassed. The results of the application to the proposed methodology is shown in “Methodology” section