凯莱公式:
spanning_trees_num( G ) = spanning_trees_num( G - e ) + spanning_trees_num( G · e )
矩阵树定理:
G 对应的拉普拉斯矩阵(度矩阵 - 邻接矩阵)L( G )
删除任意一行一列得到的行列式的值det( L*( G ) )
即生成树的个数,即spanning_trees_num( G ) = det( L*( G ) )
证:
归纳假设 spanning_trees_num( G - e ) = det( L*( G - e ) )
spanning_trees_num( G · e ) = det( L*( G · e ) )
目的就是只需要证 det( L*( G · e ) ) + det( L*( G - e ) ) = det( L*( G ) )
可以重标记顶点,取第一个点和第二个点之间的边做实验,可以得到如下矩阵
(易证删除第一行第一列的行列式值加起来相等,不废话,盯图看一会儿就明了了)
import networkx as nx
import matplotlib.pyplot as mpl
graph = nx.Graph()
graph.add_nodes_from( range( 4 ) )
graph.add_edge( 0, 1 )
graph.add_edge( 1, 2 )
graph.add_edge( 2, 3 )
graph.add_edge( 3, 0 )
graph.add_edge( 0, 2 )
nx.draw( graph )
mpl.show()
生成图:
其 det( L*( G ) ):
import numpy as np
M = nx.to_numpy_matrix( graph )
A = np.asarray( M )
I = np.identity( A.shape[0] )
D = I * np.sum( A, axis = 1 )
L = D - A
L_ = L[ 1:, 1: ]
print np.linalg.det( L_ )
为 8