Spectral Clustering
Graph Laplacian eigenvectors reveal manifold structure
Settings
Dataset
Concentric Rings
Two Moons
Spirals
Gaussian Blobs
Pinwheel
Points
200
Clusters k
2
σ (kernel width)
0.30
kNN neighbors
10
Generate Data
Run Spectral Clustering
Ready
Spectral clustering
: build a kNN graph, compute the graph Laplacian L=D-W, embed points using k smallest eigenvectors, then k-means in eigenspace. Captures non-convex clusters invisible to standard methods.