Scribd
Upload
6 views

RA#1

Uploaded by

usasua1112
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
6 views

RA#1

Uploaded by

usasua1112
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

Summary : The Matrix-Tree Theorem

Cayley’s Formula: Number of labeled trees on n vertices is nn−2 .


Proof of Cayley’s Formula (Monomial method): We want to show that treesT m(T ) − (x1 + · · · + xn )n−2 is
P
identically zero. Proof by induction:
- Base case: True for n = 2
- Inductive step: Show xi is a factor of treesT m(T ) − (x1 + · · · + xn )n−2 for all 1 ≤ i ≤ n.
P
Set x1 = · · · = xn = 1, then all trees have weight 1 and get total weight nn−2 .
Spanning Tree: Tree including all vertices of a graph G and is a subgraph of G.
Definition (Adjacency Matrix). Matrix A of graph G: n × n matrix, Aij = number of edges connecting i to j.
(
deg(i) i=j
Definition (Laplacian Matrix). L = D − A, D is a diagonal matrix. Lij =
−# edges i ↔ j i ̸= j
Theorem (Matrix-Tree Theorem). For any 1 ≤ i ≤ n, det(Li ) (L with ith row/column removed) = number of spanning
trees of G.
 
(n − 1) −1 ··· −1
 −1 (n − 1) · · · −1 
Example Laplacian matrix for Kn : L =  .
 
.. .. .. 
 .. . . . 
−1 −1 · · · (n − 1)
Proof of Cayley’s Formula using Matrix-Tree Theorem:
Key facts: 1) det(A) = product of eigenvalues 2) tr(A) = sum of eigenvalues 3) If λ is eigenvalue of A, λ + c is eigenvalue
of A + cI
Li − nI (all −1s) has rank 1, so all eigenvalues = 0(except for last one = 1 − n). Adding n to each eigenvalue: Li
eigenvalues are 1, n, . . . , n. Thus, det(Li ) = nn−2 .
Directed Graphs and Generalized Theorem
in-trees and out-trees rooted at v is a spanning tree of directed graph G with at least two vertices( except that all edges
away from/toward to v).
( (
indeg(i) i = j outdeg(i) i=j
Definition (Directed Laplacian Matrices). (Lin )ij = (Lout )ij =
−# edges i → j i ̸= j −# edges i → j i ̸= j
Theorem (Directed Matrix-Tree Theorem). For directed G, vertex v:
# out-trees at v: (−1)i+v det(Lin )iv for any 1 ≤i≤n

# in-trees at v: (−1)v+i det(Lout )vi for any 1 ≤i≤n


Proof of Directed Matrix-Tree Theorem
Induct on the number of directed edges not starting from v. Let Inv (G) be the number of in-trees of G rooted at v.
Base case: All edges out of v ⇒ no in-trees, Lout has all zero entries except row v, any cofactor (Lout )vi = 0.
Inductive step: For edge e : i → j (i ̸= v)
• G1 : G with e removed
• G2 : G with all edges from i except e removed
• Inv (G) = Inv (G1 ) + Inv (G2 )
Compare Laplacians:
• Lout (G1 ): In row i, decrease ai by 1, increase aj by 1.
• Lout (G2 ): Row i has 1 (col i), -1 (col j), 0 elsewhere
Sum of ith rows of Lout (G1 ) and Lout (G2 ) add up to the ith row of Lout (G).
By linearity of determinants (whenever not removing ith row):

(Lout (G))vi = (Lout (G1 ))vi + (Lout (G2 ))vi

Induction complete: RHS counts the number of in-trees for G1 and G2 separately.

You might also like