13.11.2014 Views

proofs

proofs

proofs

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

248 Communicating without errors<br />

⎛<br />

A =<br />

⎜<br />

⎝<br />

0 1 0 0 1<br />

1 0 1 0 0<br />

0 1 0 1 0<br />

0 0 1 0 1<br />

1 0 0 1 0<br />

⎞<br />

⎟<br />

⎠<br />

The adjacency matrix for the 5-cycle C 5<br />

Let us carry our discussion a little further. We see from (8) that the larger σ T<br />

is for a representation of G, the better a bound for Θ(G) we will get. Here<br />

is a method that gives us an orthonormal representation for any graph G.<br />

To G = (V, E) we associate the adjacency matrix A = (a ij ), which is<br />

defined as follows: Let V = {v 1 , . . .,v m }, then we set<br />

{ 1 if vi v<br />

a ij :=<br />

j ∈ E<br />

0 otherwise.<br />

A is a real symmetric matrix with 0’s in the main diagonal.<br />

Now we need two facts from linear algebra. First, as a symmetric matrix,<br />

A has m real eigenvalues λ 1 ≥ λ 2 ≥ . . . ≥ λ m (some of which may<br />

be equal), and the sum of the eigenvalues equals the sum of the diagonal<br />

entries of A, that is, 0. Hence the smallest eigenvalue must be negative<br />

(except in the trivial case when G has no edges). Let p = |λ m | = −λ m be<br />

the absolute value of the smallest eigenvalue, and consider the matrix<br />

M := I + 1 p A,<br />

where I denotes the (m × m)-identity matrix. This M has the eigenvalues<br />

1+ λ1 λ2<br />

p<br />

≥ 1+<br />

p ≥ . . . ≥ 1+ λm p<br />

= 0. Now we quote the second result (the<br />

principal axis theorem of linear algebra): If M = (m ij ) is a real symmetric<br />

matrix with all eigenvalues ≥ 0, then there are vectors v (1) , . . .,v (m) ∈ R s<br />

for s = rank(M), such that<br />

m ij = 〈v (i) , v (j) 〉<br />

In particular, for M = I + 1 pA we obtain<br />

(1 ≤ i, j ≤ m).<br />

〈v (i) , v (i) 〉 = m ii = 1 for all i<br />

and<br />

〈v (i) , v (j) 〉 = 1 p a ij for i ≠ j.<br />

Since a ij = 0 whenever v i v j ∉ E, we see that the vectors v (1) , . . .,v (m)<br />

form indeed an orthonormal representation of G.<br />

Let us, finally, apply this construction to the m-cycles C m for odd m ≥ 5.<br />

Here one easily computes p = |λ min | = 2 cos π m<br />

(see the box). Every<br />

row of the adjacency matrix contains two 1’s, implying that every row of<br />

the matrix M sums to 1 + 2 p . For the representation {v(1) , . . . ,v (m) } this<br />

means<br />

〈v (i) , v (1) + . . . + v (m) 〉 = 1 + 2 p = 1 + 1<br />

cos π m<br />

and hence<br />

〈v (i) , u〉 = 1 m (1 + (cos π m )−1 ) = σ<br />

for all i. We can therefore apply our main result (8) and conclude<br />

Θ(C m ) ≤<br />

m<br />

1 + (cos π m )−1 (for m ≥ 5 odd). (9)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!