Laplacian matrix

Laplacian matrix

From Wikipedia, the free encyclopedia

https://en.wikipedia.org/wiki/Laplacian_matrix

In the mathematical field of graph theory, the Laplacian matrix, sometimes called admittance matrix, Kirchhoff matrix or discrete Laplacian, is a matrixrepresentation of a graph. The Laplacian matrix can be used to find many useful properties of graph. Together with Kirchhoff's theorem, it can be used to calculate the number of spanning trees for a given graph. The sparsest cut of a graph can be approximated through the second eigenvalue of its Laplacian by Cheeger's inequality .

Definition[edit]

Given a simple graph G with n vertices, its Laplacian matrix {\displaystyle L_{n\times n}} is defined as:^[1]

{\displaystyle L=D-A,}

where D is the degree matrix and A is the adjacency matrix of the graph. In the case of directed graphs, either the indegree or outdegree might be used, depending on the application.

The elements of {\displaystyle L} are given by

{\displaystyle L_{i,j}:={\begin{cases}\deg(v_{i})&{\mbox{if}}\ i=j\\-1&{\mbox{if}}\ i\neq j\ {\mbox{and}}\ v_{i}{\mbox{ is adjacent to }}v_{j}\\0&{\mbox{otherwise}}\end{cases}}}

where deg(v_i) is degree of the vertex i.

The symmetric normalized Laplacian matrix is defined as:^[1]

{\displaystyle L^{\text{sym}}:=D^{-1/2}LD^{-1/2}=I-D^{-1/2}AD^{-1/2}} ,

The elements of {\displaystyle L^{\text{sym}}} are given by

{\displaystyle L_{i,j}^{\text{sym}}:={\begin{cases}1&{\mbox{if}}\ i=j\ {\mbox{and}}\ \deg(v_{i})\neq 0\\-{\frac {1}{\sqrt {\deg(v_{i})\deg(v_{j})}}}&{\mbox{if}}\ i\neq j\ {\mbox{and}}\ v_{i}{\mbox{ is adjacent to }}v_{j}\\0&{\mbox{otherwise}}.\end{cases}}}

The random-walk normalized Laplacian matrix is defined as:

{\displaystyle L^{\text{rw}}:=D^{-1}L=I-D^{-1}A}

The elements of {\displaystyle L^{\text{rw}}} are given by

{\displaystyle L_{i,j}^{\text{rw}}:={\begin{cases}1&{\mbox{if}}\ i=j\ {\mbox{and}}\ \deg(v_{i})\neq 0\\-{\frac {1}{\deg(v_{i})}}&{\mbox{if}}\ i\neq j\ {\mbox{and}}\ v_{i}{\mbox{ is adjacent to }}v_{j}\\0&{\mbox{otherwise}}.\end{cases}}}

Example[edit]

Here is a simple example of a labeled graph and its Laplacian matrix.

Labeled graph	Degree matrix	Adjacency matrix	Laplacian matrix
	{\displaystyle \left({\begin{array}{rrrrrr}2&0&0&0&0&0\\0&3&0&0&0&0\\0&0&2&0&0&0\\0&0&0&3&0&0\\0&0&0&0&3&0\\0&0&0&0&0&1\\\end{array}}\right)}	{\displaystyle \left({\begin{array}{rrrrrr}0&1&0&0&1&0\\1&0&1&0&1&0\\0&1&0&1&0&0\\0&0&1&0&1&1\\1&1&0&1&0&0\\0&0&0&1&0&0\\\end{array}}\right)}	{\displaystyle \left({\begin{array}{rrrrrr}2&-1&0&0&-1&0\\-1&3&-1&0&-1&0\\0&-1&2&-1&0&0\\0&0&-1&3&-1&-1\\-1&-1&0&-1&3&0\\0&0&0&-1&0&1\\\end{array}}\right)}

Properties[edit]

For an (undirected) graph G and its Laplacian matrix L with eigenvalues {\displaystyle \lambda _{0}\leq \lambda _{1}\leq \cdots \leq \lambda _{n-1}}:

• L is symmetric.
• L is positive-semidefinite (that is {\displaystyle \lambda _{i}\geq 0} for all i). This is verified in the incidence matrix section (below). This can also be seen from the fact that the Laplacian is symmetric and diagonally dominant.
• L is an M-matrix (its off-diagonal entries are nonpositive, yet the real parts of its eigenvalues are nonnegative).
• Every row sum and column sum of L is zero. Indeed, in the sum, the degree of the vertex is summed with a "-1" for each neighbor.
• In consequence, {\displaystyle \lambda _{0}=0}, because the vector {\displaystyle \mathbf {v} _{0}=(1,1,\dots ,1)} satisfies {\displaystyle L\mathbf {v} _{0}=\mathbf {0} .}
• The number of times 0 appears as an eigenvalue in the Laplacian is the number of connected components in the graph.
• The smallest non-zero eigenvalue of L is called the spectral gap.
• The second smallest eigenvalue of L is the algebraic connectivity (or Fiedler value) of G.
• The Laplacian is an operator on the n-dimensional vector space of functions f : V → {\displaystyle \mathbb {R} }, where V is the vertex set of G, and n = |V|.
• When G is k-regular, the normalized Laplacian is: {\displaystyle {\mathcal {L}}={\tfrac {1}{k}}L=I-{\tfrac {1}{k}}A}, where A is the adjacency matrix and I is an identity matrix.
• For a graph with multiple connected components, L is a block diagonal matrix, where each block is the respective Laplacian matrix for each component, possibly after reordering the vertices (i.e. L is permutation-similar to a block diagonal matrix).
• Laplacian matrix is singular.