Fig. 8
From: Unsupervised evaluation of multiple node ranks by reconstructing local structures
Example of evaluating the ability of two unsupervised node rank quality measures (conductance and HopAUC) to compare three ranking algorithms in the same way as supervised AUC does. As a final step, we need to select the best of the unsupervised measures. Experiments in this work span 6 unsupervised evaluation measures, 72 ranking algorithms, are conducted for 25 experiments settings -which involve multiple metadata groups- and use both AUC and DCG for supervised comparison