Lab 23 Interactive: Spectral Graph Theory

Build a graph, compute $A$, $D$, and $L=D-A$, inspect eigenvalues, count walks with $A^k$, and use the Fiedler vector for graph partitioning.

1. Choose a graph

Vertices are labeled $1,2,\ldots,n$. Click two vertices in the drawing to toggle an edge.

2. Graph drawing

Node color shows the Fiedler vector when the graph is connected. Negative and positive signs suggest a two-way spectral partition.

3. Graph matrices

4. Spectrum and connected components

5. Walk counting

The theorem says that $(A^k)_{ij}$ counts walks of length $k$ from vertex $i$ to vertex $j$.

6. Laplacian energy

Use the signal $x_i=i$ on vertices. The tool compares $x^TLx$ with $\sum_{\{i,j\}\in E}(x_i-x_j)^2$.

7. Independent-study tasks with answers

Task A. Select the disconnected graph. How many zero Laplacian eigenvalues do you see?
Answer: The number of zero eigenvalues equals the number of connected components.
Task B. Select two triangles + bridge. What partition is suggested by the Fiedler vector?
Answer: The two triangles should be separated into two clusters.
Task C. Select $P_6$. What does $(A^2)_{1,3}$ count?
Answer: The number of length-2 walks from vertex 1 to vertex 3.
Similar practice. Add or remove one bridge edge. How does the Fiedler value $\lambda_2$ change?
Answer: Stronger or more numerous connections between two parts usually increase $\lambda_2$.

8. Workspace