Sie sind auf Seite 1von 7

TAKE HOME EXAM

MICHAEL TWITO
July 26, 2018

1 Claims
Claim: Find a two non-zero eigenvalues of K1,n
P
Proof. Note that λx(u) = v∼x(v) denote u as the lone vertex in his partition,
so
λx(u) = x(v1 ) + ... + x(vn )
and
λx(vi ) = x(u), ∀i ∈ [n]
then
n
λx(u) = x(u)
λ
√ √
and so λ ∈ {− n, n}.

Claim: Let G be a graph then λn (G) < 0.


Proof. By Lemma 3 [1] TrA = 0 and by [2] we note that the trace is equal
to the sum of eigenvalues, then we must have λn < 0 to nullify the sum.

Claim: Let H be a connected spanning subgraph of a connected graph G


then λ1 (H) ≤ λ1 (G)
Proof. As H is a spanning subgraph, it must have that we omit atleast one
1, say k, l and we have that in his adj. matrix Ak,l and Al,k = 0. so it will
nullify the xk xl ,so we have that

λ1 (G) = xT A(G)x ≥ z t A(H)z = λ1 (H)

for x ∈ εA(G) (λ1 ) and z ∈ εA(H) (λ1 )

1
Claim: Let H be a subgraph of a connected graph G then λ1 (H) ≤ λ1 (G),Further
if H is a proper subgraph then λ1 (H) < λ1 (G)
Proof. Let x be an eigenvector of length 1 of the adj. matrix of H s.t it has
only non-negative enteries by Perron-Forbenius.Then

λ1 (H) = xA (H)x ≤ xT A(G)x ≤ max z T A(G)z = λ1 (G)

. If H has less vertices than G, then we complete x with 0’s in the remaining
vertices and we denote the obtained vector with x too in order to make sense
for xT A(G)x.
Assume for contradiction that λ1 (H) = λ1 (G). Then we have equality every-
where in the above computation. In particular x xT Ax = λ1 (G). This means
that x is eigenvector of A(G) too. Since G is connected x must be a (or rather
”the”) vector with only positive entries. But then xT A(H)x ≤ xT A(G)x, a
contradiction.

Claim: Let ∆ be the maximum degree, and let d denote the average degree
Then √
max( ∆, d) ≤ λ1 ≤ ∆
[3]
Proof. Let v = (1, ..., 1). Then

v T Av 2e(G)
λ1 ≥ 2
= =d
||v|| n
. If the largest degree is ∆ then G contains K1,∆ as a subgraph. Hence

λ1 (G) ≥ λ1 (K1,∆ ) = ∆

Finally, let x inεA (λ1 ). Let xi be the entry with largest absolute value. Then
X X X
|λ1 ||xi | = | ai,j xj | ≤ ai,j |xj | ≤ ai,j |xi | ≤ ∆|xi |
j j j

Hence λ1 ≤ ∆.

Claim: Let G be a connected d-regular graph on n-vertex and let λ(G) :=


max2≤i≤n |λi |, Then
s
d|X||Y | |X| |Y |
|eG (X, Y ) − | ≤ λ(G) |X||Y |(1 − )(1 − )
n n n

2
Expander Mixing Lemma:[4]

Let G be connected d-regular n-vertex graph and let X, Y ⊆ V (G) with


|X| = αn and |Y | = βn Then
q
|eG (X, Y ) − αβdn| ≤ λ(G)n (α − α2 )(β − β 2 )
Now we can prove our claim.
Proof. Subtitute α and β with |X| and |Y | we get that
s
d|X||Y | |X| |X|2 |Y | |Y |2
|eG (X, Y ) − | ≤ λ(G)n − 2 )( − 2 )
n n n n n
Rearrenging we get that
s
d|X||Y | n |X| |Y |
|eG (X, Y ) − | ≤ λ(G) |X||Y |(1 − )(1 − )
n n n n
and the claim follows.

Claim: Let G and λ(G) be as the last claim then



λ(G) ≥ d(1 − o(1))
Alon-Boppana Theorem:

let c be a constant and ∆ be the diameter of G. Then


√ c
λ2 (G) ≥ 2 d − 1(1 − 2 )

Proof. The simple bound comes from a trace method. we have
trace(A2 ) = λ2i
X

by using one definition of the trace and


trace(A2 ) = (A2 )v,v ≥ dn
X

using the other definition and observing that (A2 )v,v counts the paths that
go from v to v in two steps, of which there are at least d: follow an edge to
a neighbor of v, then follow the same edge back.so we have
dn ≤ d2 + λ2i
X

i=2,...,n

3
And so

s
n−d
maxi=2,...,n |λi | ≥ d
n−1

The condition d ≤ n(1 − ) is necessary to get lower bound of Ω( d); in the
clique, for example we have d = n − 1 and λ2 = λn = −1 A trace argument
does not give us a lower bound on λ2 , and in face it is possible to have λ2 = 0
and d = n/2,for example the complete bipartite graph. √
If the diameter of G is at least 4, it is easy to see that λ2 ≥ d. Let a,√b be
two vertices at distance 4. Define a vector x√as follows:xa = 1, xv = 2/ d if
v is a neighbor of a, xb = −1 and xv =√−1/ d if v is a neighbor of b. Then
we see that ||x||2 = 4, that xT Ax ≥ 4 d(because√ there are 2d edges, each
P
counterd twice, that give a contribution of 1 d to u,v Au, vxu xv ) and that
x is orthogonal to(1,...,1).

Claim: Let M be an irreducible Markov chain with state space Ω and tran-
sition matrix P, when M have a period d ¿ we may write
d−1
]
Ω= Ci
i=0

s.t for every i ∈ Ck ,k =0,...,d-1.


X
Pi,j = 1
j∈Ck+1

Proof. As M is irreducible we may divide Ω to disjoint equivalence classes


s.t when we fix i every Ck contains the j states reachable at the time of k,
assume toward a contradiction that there is some j that is reachable from
i at the time k < d and time l < d when k 6≡ l(mod d) Then j ∈ Ck and
j ∈ Cl which cannot happent because we have a period of d and j cannont
revisit himself by the time of l − k. As we sum over a row i and Ck contains
P
all of the reachable states j∈Ck+1 Pi,j = 1 as k is the time we may write
P k k
j Pi,j = 1 and the claim follows.

Claim: Let M be a markov chain with transition matrix P and state space
Ω For i, j ∈ Ω and n ∈ Z+ define
(n)
fi,j := Pr[Tj = n|X0 = i]

4
Then ∀n ≥ 1
n
(n) X (m) (n−m)
Pi,j = fi,j Pj,j
m=1

(n)
Proof. Observe that fi,j is he first time we hit j at the time of n starting from
(n)
i, Then we may write it as fi,j = Pr[Xn = j, Xn−1 6= j, ...X1 6= j|X0 = i]
subtitue in our sum
n
(n) X
Pi,j = Pr[Xm = j, Xn−1 6= j, ...X1 6= j|X0 = i]Pr[Xn−m = j|X0 = j]
m=1

Which is take m steps from i to j then ’spend’ the rest n − m steps on a


walk from j to j.

Claim: Let j ∈ Ω be transient then


(n)
lim Pi,j = 0
n→∞

Proof. By Lemma 14 in [5] we have that if j is transient then


t
lim Pj,j =0
n→∞

then X
lim Pj,j = 0
n→∞
j

so we may write
n ∞ X
n
n
X (m) (n−m) X (m) (n−m)
lim Pi,i = fi,j Pj,j = fi,j Pj,j
n→∞
m=1 n=1 m=1

as j is transient and the sum of f 0 s is actually the Nj defined in [5] with


ExNj and by Lemma 14 it is < ∞ it holds that ∞
P
n=1 fi,j = c Then

n
X (n−m)
lim c
n→∞
Pj,j =0
m=1

Claim: Every graph G on an even number of vertices n having minumum


degree > n/2 has perfect maching.

5
Proof. Arbitrarily divide the vertices into two sets A and B of equal size. In
each set, find a vertex with a minimal number of edges connecting it to the
other set, and swap the two minimal vertices. Repeat this until the swap
would no longer increase the total number of edges connecting the two sets.
(Since there is a finite number of edges, this must happen at some point.)
Denoting the two vertices whose swap would no longer increase the number of
connections by vA and vB and the numbers of their edges within and between
the sets by vAA , vAB , vBA and vBB , and counting the number of connections
between the sets before and after the swap, we have vAB + vBA ≥ vAA + vBB .
(There might be an edge between vA and vB, but that works in favour of
the inequality.) It follows that
1 1 1
vAB +vBA ≥ (vAB +vBA +vAA +vBB ) = (degvA +degvB ) ≥ (n/2+n/2) = n/2
2 2 2
. Now consider a subset X of A. If |X| ≤ vAB , then since each element of A
has edges to at least vAB elements of B, X has at least as many neighbours
in B as elements. If |X| ≥ vAB , then since each element of B has edges to
at least vBA elements of A and vAB + vBA ≥ n/2 = |A|, at least one of these
edges must lead to X, so X has n/2 neighbours in B, and thus at least as
many as it has elements. Thus the premise of Hall’s theorem is fulfilled, and
the bipartite graph induced on A and B must contain a perfect matching,
which is also a perfect matching in G.[6]

Claim: Every bipartite graph on an even number of vertices n having both


its colour classes of size n/2 and minimum degree > n/4 has Hamilton cycle.
Proof. As the colour classes in size of n/2 it must be that we have only two
of them, this two will be our partitions,so G is Gn/2,n/2 which is the condition
for Hamilonicity in bipartite [7] we have that G is Hamiltonian.

2 References
[1]Information from the spectrum of the adj. matrix(Elad Aigner-Horev)

[2]https://math.stackexchange.com/questions/546155/proof-that-the-trace-of-
a-matrix-is-the-sum-of-its-eigenvalues

[3]Spectral Graph Theory (Cvsikvari,M.I.T)

6
[4] N. Alon and F. R. K. Chung, Explicit construction of linear sized tol-
erant networks, Proceedings of the First Japan Conference on Graph Theory
and Applications (Hakone, 1986), vol. 72, 1988, pp. 15–19. MR 975519

[5]Classifying states of Markov chains(Elad Aigner-Horev)

[6] https://math.stackexchange.com/questions/56063/minimum-degree-of-a-
graph-and-existence-of-perfect-matching

[7] https://proofwiki.org/wiki/Condition for Bipartite Graph to be Hamiltonian

Das könnte Ihnen auch gefallen