Fig. 4: Progression of DBSCAN clusters with varying neighborhood radius parameters ε. | npj Computational Materials

Fig. 4: Progression of DBSCAN clusters with varying neighborhood radius parameters ε.

From: Spectral operator representations

Fig. 4

The number of materials in each initial cluster is shown in parentheses. As the neighborhood radius is increased, clusters merge, agreeing with the MC3D's structure-based grouping for values from 2 × 10−2 to ~5 × 10−2, after which meaningfully distinct clusters combine.

Back to article page