Spectral Clustering — Graph Laplacian

The graph Laplacian L = D − A has a spectrum that encodes community structure. The Fiedler vector (second smallest eigenvector of L) provides the optimal graph cut: sign(v₂) partitions the graph into two clusters. Watch how the Fiedler vector perfectly separates communities even with cross-cluster noise edges.

12
3
0.70
0.05
3
N = 0 Edges = 0 λ₂ = 0 Conductance = 0