@ -32,7 +32,7 @@ What kinds of matrices can we analyze using spectral graph theory?
In particular, $$\lambda_2$$, the second smallest eigenvalue of $$L$$, is already fascinating and studying it will let us make big strides in understanding graph clustering. By the theory of Rayleigh quotients, we have that $$\lambda_2 = \min_{x: x^T w_1 = 0} \frac{x^T L x}{x^T x}$$ where $$w_1$$ is the eigenvector corresponding to eigenvalue $$\lambda_1$$; in other words, we minimize the objective in the subspace of vectors orthogonal to the first eigenvector in order to find the second eigenvector (remember that $$L$$ is symmetric and thus has an orthogonal basis of eigenvalues). On a high level, Rayleigh quotients frame the eigenvector search as an optimization problem, letting us bring optimization techniques to bear. Note that the objective value does not depend on the magnitude of $$x$$, so we can constrain its magnitude to be 1. Note additionally that we know that the first eigenvector of $$L$$ is the all-ones vector with eigenvalue 0, so saying that $$x$$ is orthogonal to this vector is equivalent to saying that $$\sum_i x_i = 0$$.
In particular, $$\lambda_2$$, the second smallest eigenvalue of $$L$$, is already fascinating and studying it will let us make big strides in understanding graph clustering. By the theory of Rayleigh quotients, we have that $$\lambda_2 = \min_{x: x^T w_1 = 0} \frac{x^T L x}{x^T x}$$ where $$w_1$$ is the eigenvector corresponding to eigenvalue $$\lambda_1$$; in other words, we minimize the objective in the subspace of vectors orthogonal to the first eigenvector in order to find the second eigenvector (remember that $$L$$ is symmetric and thus has an orthogonal basis of eigenvalues). On a high level, Rayleigh quotients frame the eigenvector search as an optimization problem, letting us bring optimization techniques to bear. Note that the objective value does not depend on the magnitude of $$x$$, so we can constrain its magnitude to be 1. Note additionally that we know that the first eigenvector of $$L$$ is the all-ones vector with eigenvalue 0, so saying that $$x$$ is orthogonal to this vector is equivalent to saying that $$\sum_i x_i = 0$$.
Using these properties and the definition of $$L$$, we can write out a more concrete formula for $$\lambda_2$$: $$\lambda_2 = \min_x \frac{\sum_{(i, j) \in E} (x_i - x_j)^2}{\sum_i x_i^2}$$, subject to the constraint $$\sum_i x_i = 0$$. If we additionally constrain $$x$$ to have unit length, the objective turns into simply $$\min_x \frac{\sum_{(i, j) \in E} (x_i - x_j)^2}$$.
Using these properties and the definition of $$L$$, we can write out a more concrete formula for $$\lambda_2$$: $$\lambda_2 = \min_x \frac{\sum_{(i, j) \in E} (x_i - x_j)^2}{\sum_i x_i^2}$$, subject to the constraint $$\sum_i x_i = 0$$. If we additionally constrain $$x$$ to have unit length, the objective turns into simply $$\min_x \sum_{(i, j) \in E} (x_i - x_j)^2$$.
How does $$\lambda_2$$ relate to our original objective of finding a best partition of our graph? Let's express our partition $$(A, B)$$ as a vector $$y$$ defined by $$y_i = 1$$ if $$i \in A$$ and $$y_i = -1$$ if $$i \in B$$. Instead of using the conductance here, let's first try to minimize the cut while taking care of the problem of balancing partition sizes by enforcing that $$|A| = |B|$$ (balance size of partitions), which amounts to constraining $$\sum_i y_i = 0$$. Given this size constraint, let's minimize the cut of the partition, i.e. find $$y$$ that minimizes $$\sum_{(i, j) \in E} (y_i - y_j)^2$$. Note that the entries of $$y$$ must be $$+1$$ or $$-1$$, which has the consequence that the length of $$y$$ is fixed. *This optimization problem looks a lot like the definition of $$\lambda_2$$!* Indeed, by our findings above we have that this objective is minimized by $$\lambda_2$$ of our Laplacian, and the optimal clustering $$y$$ is given by its corresponding eigenvector, known as the **Fiedler vector**.
How does $$\lambda_2$$ relate to our original objective of finding a best partition of our graph? Let's express our partition $$(A, B)$$ as a vector $$y$$ defined by $$y_i = 1$$ if $$i \in A$$ and $$y_i = -1$$ if $$i \in B$$. Instead of using the conductance here, let's first try to minimize the cut while taking care of the problem of balancing partition sizes by enforcing that $$|A| = |B|$$ (balance size of partitions), which amounts to constraining $$\sum_i y_i = 0$$. Given this size constraint, let's minimize the cut of the partition, i.e. find $$y$$ that minimizes $$\sum_{(i, j) \in E} (y_i - y_j)^2$$. Note that the entries of $$y$$ must be $$+1$$ or $$-1$$, which has the consequence that the length of $$y$$ is fixed. *This optimization problem looks a lot like the definition of $$\lambda_2$$!* Indeed, by our findings above we have that this objective is minimized by $$\lambda_2$$ of our Laplacian, and the optimal clustering $$y$$ is given by its corresponding eigenvector, known as the **Fiedler vector**.