Extended Data Fig. 9: Prediction labeling stability.
From: Machine learning enables scalable and systematic hierarchical virus taxonomy
Line plots of adjusted rand index (ARI) and normalized mutual information (NMI) of a set of prediction labels. Colors indicate rank, dashed lines exclude singleton groups from the analysis, whereas solid lines include singletons. “Successive” plots are when ARI/NMI is calculated between fractions, “cumulative” is ARI/NMI calculated against the final (100%) fraction. High ARI and NMI indicate high agreement between genome predictions (that is labels) as more data is added, where an ARI and NMI of 1.00 indicate perfect agreement between labels between datasets.
