🔮 SPECTRAL CLUSTERING
Number of points:
80
Number of clusters k:
3
σ (kernel bandwidth):
0.15
kNN neighbors:
8
Generate New Data
Run Spectral Clustering
Left: data + edges
Right: Fiedler embedding
Normalized cut: Lsym = D^{-1/2} L D^{-1/2}. Eigenvectors of k smallest eigenvalues → k-means in spectral space.