Fig. 1
From: Semi-supervised graph labelling reveals increasing partisanship in the United States Congress
Clockwise from top right: histogram of absorption probabilities for the 110th Senate; graph of the 110th Senate; graph of the 90th Senate; histogram of absorption probabilities for the 90th Senate. In the histograms, red bars represent Republican members and blue bars represent Democrat members. In the graphs, red nodes represent Republican members and blue nodes represent Democrat members. Graphs are visualised using the Fruchterman-Reingold force-directed layout algorithm. The 90th Senate was the most difficult to correctly label using GLaSS (F1 = 0.8571). This diffculty is reflected in the overlap of red and blue bars in the histogram (top left), and interspersal of red and blue nodes in the graph (bottom left). The isolated red node on the far right of the graph corresponds to the isolated red bar on the far right of the histogram. The 110th Senate was perfectly labelled by GLaSS (F1 score = 1), as reflected by the clear separation of red and blue bars in the histogram (top right). Also of note is the clearer separation of red and blue nodes within the graph of the 110th Senate (bottom right), suggesting a more obvious community structure