NumXL: Difference between revisions
en>FrescoBot m Bot: link syntax/spacing and minor changes |
en>Spiderfinancial m linked feature with an existing wiki page |
||
Line 1: | Line 1: | ||
A certain family of [[BCH code]]s have a particularly useful property, which is that | |||
treated as [[linear operator]]s, their [[dual basis|dual operators]] turns their input into an <math>\ell</math>-wise [[pairwise independence|independent source]]. That is, the set of vectors from the input [[vector space]] are mapped to an <math>\ell</math>-wise independent source. The proof of this fact below as the following Lemma and Corollary is useful in derandomizing the algorithm for a <math>1-2^{-\ell}</math>-approximation to [[MAXEkSAT]]. | |||
==Lemma== | |||
Let <math>C\subseteq F_2^n</math> be a linear code such that <math>C^\perp </math> has distance greater than <math> \ell +1</math>. Then <math>C</math> is an <math>\ell</math>-wise independent source. | |||
==Proof of Lemma== | |||
It is sufficient to show that given any <math>k \times l</math> matrix ''M'', where ''k'' is greater than or equal to ''l'', such that the rank of ''M'' is ''l'', for all <math>x\in F_2^k</math>, <math>xM</math> takes every value in <math>F_2^l</math> the same number of times. | |||
Since ''M'' has rank ''l'', we can write ''M'' as two matrices of the same size, <math>M_1</math> and <math>M_2</math>, where <math>M_1</math> has rank equal to ''l''. This means that <math>xM</math> can be rewritten as <math>x_1M_1 + x_2M_2</math> for some <math>x_1</math> and <math>x_2</math>. | |||
If we consider ''M'' written with respect to a basis where the first ''l'' rows are the identity matrix, then <math>x_1</math> has zeros wherever <math>M_2</math> has nonzero rows, and <math>x_2</math> has zeros wherever <math>M_1</math> has nonzero rows. | |||
Now any value ''y'', where <math>y=xM</math>, can be written as <math>x_1M_1+x_2M_2</math> for some vectors <math>x_1, x_2</math>. | |||
We can rewrite this as: | |||
<math>x_1M_1 = y - x_2M_2</math> | |||
Fixing the value of the last <math>k-l</math> coordinates of | |||
<math>x_2\in F_2^k</math> (note that there are exactly <math>2^{k-l}</math> | |||
such choices), we can rewrite this equation again as: | |||
<math>x_1M_1 = b</math> for some ''b''. | |||
Since <math>M_1</math> has rank equal to ''l'', there | |||
is exactly one solution <math>x_1</math>, so the total number of solutions is exactly <math>2^{k-l}</math>, proving the lemma. | |||
==Corollary== | |||
Recall that BCH<sub>2,''m'',''d''</sub> is an <math> [n=2^m, n-1 -\lceil {d-2}/2\rceil m, d]_2</math> linear code. | |||
Let <math>C^\perp</math> be BCH<sub>2,log ''n'',''ℓ''+1</sub>. Then <math>C</math> is an <math>\ell</math>-wise independent source of size <math>O(n^{\lfloor \ell/2 \rfloor})</math>. | |||
==Proof of Corollary== | |||
The dimension ''d'' of ''C'' is just <math>\lceil{(\ell +1 -2)/{2}}\rceil \log n +1 </math>. So <math>d = \lceil {(\ell -1)}/2\rceil \log n +1 = \lfloor \ell/2 \rfloor \log n +1</math>. | |||
So the cardinality of <math>C</math> considered as a set is just | |||
<math> 2^{d}=O(n^{\lfloor \ell/2 \rfloor})</math>, proving the Corollary. | |||
== References == | |||
[http://www.cse.buffalo.edu/~atri/courses/coding-theory/ Coding Theory notes at University at Buffalo] | |||
[http://people.csail.mit.edu/madhu/FT01/course.html/ Coding Theory notes at MIT] | |||
[[Category:Article proofs]] |
Revision as of 13:56, 7 August 2013
A certain family of BCH codes have a particularly useful property, which is that treated as linear operators, their dual operators turns their input into an -wise independent source. That is, the set of vectors from the input vector space are mapped to an -wise independent source. The proof of this fact below as the following Lemma and Corollary is useful in derandomizing the algorithm for a -approximation to MAXEkSAT.
Lemma
Let be a linear code such that has distance greater than . Then is an -wise independent source.
Proof of Lemma
It is sufficient to show that given any matrix M, where k is greater than or equal to l, such that the rank of M is l, for all , takes every value in the same number of times.
Since M has rank l, we can write M as two matrices of the same size, and , where has rank equal to l. This means that can be rewritten as for some and .
If we consider M written with respect to a basis where the first l rows are the identity matrix, then has zeros wherever has nonzero rows, and has zeros wherever has nonzero rows.
Now any value y, where , can be written as for some vectors .
We can rewrite this as:
Fixing the value of the last coordinates of (note that there are exactly such choices), we can rewrite this equation again as:
Since has rank equal to l, there is exactly one solution , so the total number of solutions is exactly , proving the lemma.
Corollary
Recall that BCH2,m,d is an linear code.
Let be BCH2,log n,ℓ+1. Then is an -wise independent source of size .
Proof of Corollary
The dimension d of C is just . So .
So the cardinality of considered as a set is just , proving the Corollary.