Spectral Graph Laplacian

Draw a graph, compute its Laplacian eigenvalues, and visualize the Fiedler vector as node colors for graph partitioning

Graph Editor

Mode: Add Node — click canvas to add, drag to move

Eigenvalue Spectrum (click bar to select eigenvector)

Eigenvector Selector

Laplacian L = D - A

Add nodes and edges to see L

Graph Properties

Spectral Graph Theory

The graph Laplacian L = D - A encodes connectivity. Its eigenvalues are always non-negative real numbers.

lambda_1 = 0 always. The multiplicity of zero eigenvalues equals the number of connected components.

The Fiedler vector (eigenvector for lambda_2) gives the optimal graph bisection. Nodes colored red vs blue are on opposite sides of the minimum cut.

Cheeger inequality: h/2 ≤ lambda_2/2 ≤ h, where h is the edge expansion. The barbell graph has small lambda_2 (bottleneck); the complete graph has large lambda_2 (highly connected).