14.02.2013 Views

Gruber P. Convex and Discrete Geometry

Gruber P. Convex and Discrete Geometry

Gruber P. Convex and Discrete Geometry

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.

428 <strong>Geometry</strong> of Numbers<br />

Examples of Codes<br />

We present three codes which are important for packing of balls:<br />

Our first example is the [d, d − 1, 2]-even weight code consisting of all words<br />

with an even number of 1s.<br />

A Hadamard matrix H is a d × d matrix with entries ±1 such that HH T = dI,<br />

where I is the d × d unit matrix. It is known that Hadamard matrices can exist<br />

only for d = 1, 2 <strong>and</strong> multiples of 4, but whether there are Hadamard matrices for<br />

all multiples of 4 is an open question. The following inductive construction yields<br />

special Hadamard matrices for powers of 2:<br />

H1 = (1), H2 =<br />

� H1 H1<br />

H1 −H1<br />

�<br />

,..., H2k =<br />

� Hk Hk<br />

Hk −Hk<br />

�<br />

,...<br />

Observe that the first row <strong>and</strong> column of H8 contain only 1s. Now consider the matrix<br />

�H8 by replacing the 1s by 0s <strong>and</strong> the −1s by 1s. Delete the first column of �H8. The<br />

rows of the remaining matrix then form a (7, 8, 4)-code. The code-words form the<br />

vertices of a regular simplex inscribed into the unit cube in E7 .<br />

The Golay code G24 is a (24, 212 , 8)-code which is defined as follows: Let G12<br />

be the 12 × 12 matrix (cik), where<br />

⎧<br />

⎨ 0ifi = k = 1ori, k ≥ 2 <strong>and</strong> i + k − 4isa<br />

cik = quadratic residue mod 11,<br />

⎩<br />

1 otherwise,<br />

<strong>and</strong> let I12 be the 12 × 12 unit matrix. Then G24 is the 12-dimensional subspace of<br />

F 24 spanned by the rows of the matrix (I12, G12).<br />

Construction of Ball Packings by Means of Codes<br />

Leech <strong>and</strong> Sloane [638] specified three basic constructions of packings of balls using<br />

error correcting binary codes. We describe two of these, constructions A <strong>and</strong> B.<br />

Theorem 29.4. Construction A: Let C be a (d, M, m)-code. We consider C as a<br />

subset of the set of vertices of the unit cube in E d . Let<br />

T = C + 2 Z d �<br />

,ρ= min 1,<br />

Then {ρ B d + t : t ∈ T } is a packing of density<br />

If C is linear, then T is a lattice.<br />

Mρd V (B d )<br />

2d .<br />

√ m<br />

2<br />

�<br />

.

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

Saved successfully!

Ooh no, something went wrong!