Gallery: optimal structural-entropy partitions

The SE-optimal flat partition (minimizing 2D structural entropy at free k, via selib.optimal_2d) on classical graphs. Node colour = community. ← benchmark · SE tutorial · repo

Each graph is partitioned by minimizing its structural entropy — no target number of clusters is given; the resolution is chosen by the objective. The calculator behind these numbers:

import selib, networkx as nx
G = nx.karate_club_graph()
selib.se_report(G)
# {'n': 34, 'm': 78, 'se_1d': ..., 'se_2d_optimal': ..., 'num_communities': ..., ...}
labels, h2 = selib.optimal_2d(G)        # SE-optimal partition + its 2D-SE
selib.structural_entropy(G, labels)     # 2D-SE of any partition
selib.structural_entropy(G, dim=1)      # 1D-SE (partition-free upper bound)