2
$\begingroup$

If $(X,d)$ is a metric space, $x\in X$ and $r\in\mathbb{R}$ with $r>0$ we set $$S_r(x) = \{z\in X: d(x,z) < r\}.$$ Given positive integers $k<n\in\mathbb{N}$ and the metric space $(\{0,1\}^n, d)$ where $d$ denotes the Hamming distance, is it known how to choose a subset $T\subseteq \{0,1\}^n$ such that $$\{S_k(t):t\in T\}$$ is a set of disjoint spheres that is maximal in cardinality amongst all sets of disjoint spheres with radius $k$, in $\{0,1\}^n$?

$\endgroup$
  • 7
    $\begingroup$ Keyword: coding theory $\endgroup$ – Boris Bukh Oct 10 '17 at 13:19
  • 1
    $\begingroup$ I suppose you're asking for a reasonably efficient algorithm to determine this maximum cardinality, right? $\endgroup$ – Wlodek Kuperberg Oct 10 '17 at 14:37
  • $\begingroup$ @WlodekKuperberg right - and how to select $T$? I will google what Boris Bukh suggested on that behalf and see whether it gives me an answer $\endgroup$ – Dominic van der Zypen Oct 10 '17 at 14:47
3
$\begingroup$

By sphere packing arguments you can't do better in general than the Hamming bound $$ \# T \left(1+ \binom{n}{1}+ \cdots+ \binom{n}{r-1}\right)\leq 2^n, $$ or $$ \#T \#S_{r-1}(x)\leq 2^n, $$ in your notation. $t=r-1$ is the number of errors the code can correct.

Codes achieving the above are called perfect, there are only 4 of them in the binary alphabet case:

  • the full space with $t=0,$ admittedly trivial,
  • the $n$ length repetition codes consisting of the all zero and all one vectors with $t=(n-1)/2$, which exists for all positive odd $n$
  • the family of Hamming codes with $t=1$ and $n=2^m-1$, and
  • the binary Golay code of length $n=23$ and $t=3$ which is related to many combinatorial objects.

There is also a lower bound called Gilbert Varshamov which asserts the existence of sets $T$ with size at least equal to the bound given $n$ and $r$.

Nice coding theory texts include those by MacWilliams and Sloane, Roman, Bierbrauer, Roth.

$\endgroup$
  • 2
    $\begingroup$ "only 4 of them" meaning "2 of them, and 2 infinite families". $\endgroup$ – Gerry Myerson Oct 10 '17 at 22:00

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.