拉普拉斯矩阵(Laplacian matrix) 也叫做导纳矩阵、基尔霍夫矩阵或离散拉普拉斯算子,是图论中用于表示图的一种重要矩阵。
定义
给定一个具有
n
n
n个顶点的简单图
G
=
(
V
,
E
)
G=(V, E)
G=(V,E),
V
V
V为顶点集合,
E
E
E为边集合,其拉普拉斯矩阵可定义为:
L
=
D
−
A
L=D-A
L=D−A其中
A
∈
R
n
×
n
A \in \mathbb{R}^{n \times n}
A∈Rn×n为邻接矩阵(adjacency matrix),
D
∈
R
n
×
n
D \in \mathbb{R}^{n \times n}
D∈Rn×n为度矩阵(degree matrix)。
注意,
A
A
A中的元素仅仅可能是 0 和 1,且其对角元素全为 0。
D
D
D是一个对角矩阵,其对角线上的元素计算方式为
D
i
i
=
∑
j
A
i
j
D_{i i}=\sum_{j} A_{i j}
Dii=∑jAij。
进一步地,
L
L
L的每个元素值的具体计算方式如下所示:
L
i
j
=
{
D
i
i
if
i
=
j
−
1
if
i
≠
j
and
v
i
is adjacent to
v
j
0
otherwise
L_{ij}=\left\{\begin{array}{ll}{D_{ii}} & {\text { if } i=j} \\ {-1} & {\text { if } i \neq j \text { and } v_{i} \text { is adjacent to } v_{j}} \\ {0} & {\text { otherwise }}\end{array}\right.
Lij=⎩⎨⎧Dii−10 if i=j if i=j and vi is adjacent to vj otherwise
示例
给定一个简单的图
G
=
(
V
,
E
)
G=(V, E)
G=(V,E),其示意图如下所示(例子来自wiki百科):
根据该示意图,可以获得
G
G
G对应的邻接矩阵
A
A
A,如下所示:
A
=
(
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
)
A=\left(\begin{array}{llllll}{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)
A=⎝⎜⎜⎜⎜⎜⎜⎛010010101010010100001011110100000100⎠⎟⎟⎟⎟⎟⎟⎞
根据度矩阵的计算方式,可以获得
G
G
G对应的度矩阵
D
D
D,如下所示:
D
=
(
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
)
D=\left(\begin{array}{llllll}{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)
D=⎝⎜⎜⎜⎜⎜⎜⎛200000030000002000000300000030000001⎠⎟⎟⎟⎟⎟⎟⎞
根据拉普拉斯矩阵的定义
L
=
D
−
A
L=D-A
L=D−A,可以获得
G
G
G对应的拉普拉斯矩阵
L
L
L,如下所示:
L
=
(
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
)
L=\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)
L=⎝⎜⎜⎜⎜⎜⎜⎛2−100−10−13−10−100−12−10000−13−1−1−1−10−130000−101⎠⎟⎟⎟⎟⎟⎟⎞显然,拉普拉斯矩阵都是对称的。
性质
- 拉普拉斯矩阵是半正定矩阵;
- 特征值中0出现的次数就是图连通区域的个数;
- 最小特征值是0,因为拉普拉斯矩阵每一行的和均为0;
- 最小非零特征值是图的代数连通度。
变体
除了最普通的拉普拉斯矩阵形式
L
=
D
−
A
L=D-A
L=D−A外,还具有多种常见的形式,如Symmetric normalized Laplacian和Random walk normalized Laplacian。
Symmetric normalized Laplacian的定义如下:
L
s
n
=
D
−
1
2
L
D
−
1
2
=
I
−
D
−
1
2
A
D
−
1
2
L^{sn}=D^{-\frac{1}{2}} L D^{-\frac{1}{2}}=I-D^{-\frac{1}{2}} A D^{-\frac{1}{2}}
Lsn=D−21LD−21=I−D−21AD−21
L
i
j
s
n
=
{
1
if
i
=
j
and
D
i
i
≠
0
−
1
D
i
i
D
j
j
if
i
≠
j
and
v
i
is adjacent to
v
j
0
otherwise
L_{ij}^{sn}=\left\{\begin{array}{ll}{1} & {\text { if } i=j \text { and } D_{ii} \neq 0} \\ {-\frac{1}{\sqrt{D_{ii} D_{jj}}}} & {\text { if } i \neq j \text { and } v_{i} \text { is adjacent to } v_{j}} \\ {0} & {\text { otherwise }}\end{array}\right.
Lijsn=⎩⎪⎨⎪⎧1−DiiDjj10 if i=j and Dii=0 if i=j and vi is adjacent to vj otherwise
Random walk normalized Laplacian的定义如下:
L
r
w
=
D
−
1
L
=
I
−
D
−
1
A
L^{r w}=D^{-1} L=I-D^{-1} A
Lrw=D−1L=I−D−1A
L
i
j
r
w
=
{
1
if
i
=
j
and
D
i
i
≠
0
−
1
D
i
i
if
i
≠
j
and
v
i
is adjacent to
v
j
0
otherwise
L_{ij}^{rw}=\left\{\begin{array}{ll}{1} & {\text { if } i=j \text { and }D_{ii} \neq 0} \\ {-\frac{1}{D_{ii}}} & {\text { if } i \neq j \text { and } v_{i} \text { is adjacent to } v_{j}} \\ {0} & {\text { otherwise }}\end{array}\right.
Lijrw=⎩⎨⎧1−Dii10 if i=j and Dii=0 if i=j and vi is adjacent to vj otherwise
本文参考资料
转载请注明出处:https://blog.csdn.net/qq280929090/article/details/103591577