Fig. 10From: Dynamic graphs, community detection, and Riemannian geometryCommunity similarity in interpolated proteomics network. When no threshold is applied (bottom right), the LE geodesic displays more extreme behaviour than the AI geodesic. A threshold of 0.1 (top right) gives the best performance for the geodesics, a threshold of 0.02 (top left) produces excessive variation in the LE geodesic, and a threshold of 0.5 (bottom left) produces almost piecewise constant behaviour in the geodesics. The linear interpolation produces reasonable results when no threshold is appliedBack to article page