Fig. 11
From: Dynamic graphs, community detection, and Riemannian geometry
Graph energy in interpolated proteomics network. Applying thresholds of 0.5 (bottom left), 0.1 (top right), and 0.02 (top left) produced the same kind of trends in the interpolations’ graph energy as was the case in considering the number of communities: low-energy (i.e., less connected) graphs with the AI geodesic, high energy (i.e., more connected) graphs with the linear interpolation, and graphs of varying energy with the LE geodesic. Interpolation without a threshold (bottom right) gave similar performance for the geodesics