拉普拉斯矩阵(Laplacian matrix),也称为导纳矩阵(Admittance matrix)或者基尔霍夫矩阵(Kirchohoff matrix)

图谱论(Spectral Graph Theory)基础-LMLPHP

归一化的拉普拉斯矩阵定义为

图谱论(Spectral Graph Theory)基础-LMLPHP

例子:

图谱论(Spectral Graph Theory)基础-LMLPHP

拉普拉斯矩阵性质:

(1)对称半正定矩阵

(2)最小特征值为0

图谱论(Spectral Graph Theory)基础-LMLPHP

证明:图谱论(Spectral Graph Theory)基础-LMLPHP图谱论(Spectral Graph Theory)基础-LMLPHP= (图谱论(Spectral Graph Theory)基础-LMLPHP图谱论(Spectral Graph Theory)基础-LMLPHP) * 图谱论(Spectral Graph Theory)基础-LMLPHP= 0 = 0 * 图谱论(Spectral Graph Theory)基础-LMLPHP

(3)任何一个属于实向量图谱论(Spectral Graph Theory)基础-LMLPHP,有以下式子成立

图谱论(Spectral Graph Theory)基础-LMLPHP

证明:

图谱论(Spectral Graph Theory)基础-LMLPHP

谱聚类:

图谱论(Spectral Graph Theory)基础-LMLPHP

矩阵的谱半径就是指矩阵的特征值中绝对值最大的那个。ρ(A)=max{|λi|,i=1,2,……n} 为A的谱半径.

ρ(A)≤║A║

图谱论(Spectral Graph Theory)基础-LMLPHP

图谱论(Spectral Graph Theory)基础-LMLPHP

图谱论(Spectral Graph Theory)基础-LMLPHP

图谱论(Spectral Graph Theory)基础-LMLPHP

图谱论(Spectral Graph Theory)基础-LMLPHP

图谱论(Spectral Graph Theory)基础-LMLPHP

图谱论(Spectral Graph Theory)基础-LMLPHP

图谱论(Spectral Graph Theory)基础-LMLPHP

图谱论(Spectral Graph Theory)基础-LMLPHP

当两个图的邻接矩阵有相同的特征值集时,它们被称为是谱相似的。

拉普拉斯矩阵的第二小特征值:

图谱论(Spectral Graph Theory)基础-LMLPHP

图谱论(Spectral Graph Theory)基础-LMLPHP

图谱论(Spectral Graph Theory)基础-LMLPHP

图谱论(Spectral Graph Theory)基础-LMLPHP

图谱论(Spectral Graph Theory)基础-LMLPHP

图谱论(Spectral Graph Theory)基础-LMLPHP

图谱论(Spectral Graph Theory)基础-LMLPHP

图谱论(Spectral Graph Theory)基础-LMLPHP

05-28 23:09