-
On the covering radius of small codes versus dual distance
Authors:
Louay Bazzi
Abstract:
Tietäväinen's upper and lower bounds assert that for block-length-$n$ linear codes with dual distance $d$, the covering radius $R$ is at most $\frac{n}{2}-(\frac{1}{2}-o(1))\sqrt{dn}$ and typically at least $\frac{n}{2}-Θ(\sqrt{dn\log{\frac{n}{d}}})$. The gap between those bounds on $R -\frac{n}{2}$ is an $Θ(\sqrt{\log{\frac{n}{d}}})$ factor related to the gap between the worst covering radius giv…
▽ More
Tietäväinen's upper and lower bounds assert that for block-length-$n$ linear codes with dual distance $d$, the covering radius $R$ is at most $\frac{n}{2}-(\frac{1}{2}-o(1))\sqrt{dn}$ and typically at least $\frac{n}{2}-Θ(\sqrt{dn\log{\frac{n}{d}}})$. The gap between those bounds on $R -\frac{n}{2}$ is an $Θ(\sqrt{\log{\frac{n}{d}}})$ factor related to the gap between the worst covering radius given $d$ and the sphere-covering bound. Our focus in this paper is on the case when $d = o(n)$, i.e., when the code size is subexponential and the gap is $w(1)$. We show that up to a constant, the gap can be eliminated by relaxing the covering requirement to allow for missing $o(1)$ fraction of points. Namely, if the dual distance $d = o(n)$, then for sufficiently large $d$, almost all points can be covered with radius $R\leq\frac{n}{2}-Θ(\sqrt{dn\log{\frac{n}{d}}})$. Compared to random linear codes, our bound on $R-\frac{n}{2}$ is asymptotically tight up to a factor less than $3$. We give applications to dual BCH codes. The proof builds on the author's previous work on the weight distribution of cosets of linear codes, which we simplify in this paper and extend from codes to probability distributions on $\{0,1\}^n$, thus enabling the extension of the above result to $(d-1)$-wise independent distributions.
△ Less
Submitted 25 July, 2018; v1 submitted 20 July, 2017;
originally announced July 2017.
-
On the tightness of Tietäväinen's bound for distributions with limited independence
Authors:
Louay Bazzi
Abstract:
In 1990, Tietäväinen showed that if the only information we know about a linear code is its dual distance $d$, then its covering radius $R$ is at most $\frac{n}{2}-(\frac{1}{2}-o(1))\sqrt{dn}$. While Tietäväinen's bound was later improved for large values of $d$, it is still the best known upper bound for small values including the $d = o(n)$ regime. Tietäväinen's bound holds also for $(d-1)$-wise…
▽ More
In 1990, Tietäväinen showed that if the only information we know about a linear code is its dual distance $d$, then its covering radius $R$ is at most $\frac{n}{2}-(\frac{1}{2}-o(1))\sqrt{dn}$. While Tietäväinen's bound was later improved for large values of $d$, it is still the best known upper bound for small values including the $d = o(n)$ regime. Tietäväinen's bound holds also for $(d-1)$-wise independent probability distributions on $\{0,1\}^n$, of which linear codes with dual distance $d$ are special cases. We show that Tietäväinen's bound on $R-\frac{n}{2}$ is asymptotically tight up to a factor of $2$ for $k$-wise independent distributions if $k\leq\frac{n^{1/3}}{\log^2{n}}$. Namely, we show that there exists a $k$-wise independent probability distribution $μ$ on $\{0,1\}^n$ whose covering radius is at least $\frac{n}{2}-\sqrt{kn}$. Our key technical contribution is the following lemma on low degree polynomials, which implies the existence of $μ$ by linear programming duality. We show that, for sufficiently large $k\leq\frac{n^{1/3}}{\log^2{n}}$ and for each polynomial $f(v)\in {\mathbb R}[v]$ of degree at most $k$, the expected value of $f$ with respect to the binomial distribution cannot be positive if $f(w)\leq 0$ for each integer $w$ such that $|w-n/2|\leq\sqrt{kn}$. The proof uses tools from approximation theory.
△ Less
Submitted 3 July, 2017;
originally announced July 2017.
-
LP decoding excess over symmetric channels
Authors:
Louay Bazzi,
Ibrahim Abou-Faycal
Abstract:
We consider the problem of Linear Programming (LP) decoding of binary linear codes. The LP excess lemma was introduced by the first author, B. Ghazi, and R. Urbanke (IEEE Trans. Inf. Th., 2014) as a technique to trade crossover probability for "LP excess" over the Binary Symmetric Channel. We generalize the LP excess lemma to discrete, binary-input, Memoryless, Symmetric and LLR-Bounded (MSB) chan…
▽ More
We consider the problem of Linear Programming (LP) decoding of binary linear codes. The LP excess lemma was introduced by the first author, B. Ghazi, and R. Urbanke (IEEE Trans. Inf. Th., 2014) as a technique to trade crossover probability for "LP excess" over the Binary Symmetric Channel. We generalize the LP excess lemma to discrete, binary-input, Memoryless, Symmetric and LLR-Bounded (MSB) channels. As an application, we extend a result by the first author and H. Audah (IEEE Trans. Inf. Th., 2015) on the impact of redundant checks on LP decoding to discrete MSB channels.
△ Less
Submitted 13 July, 2015;
originally announced July 2015.
-
Impact of redundant checks on the LP decoding thresholds of LDPC codes
Authors:
Louay Bazzi,
Hani Audah
Abstract:
Feldman et al.(2005) asked whether the performance of the LP decoder can be improved by adding redundant parity checks to tighten the LP relaxation. We prove that for LDPC codes, even if we include all redundant checks, asymptotically there is no gain in the LP decoder threshold on the BSC under certain conditions on the base Tanner graph. First, we show that if the graph has bounded check-degree…
▽ More
Feldman et al.(2005) asked whether the performance of the LP decoder can be improved by adding redundant parity checks to tighten the LP relaxation. We prove that for LDPC codes, even if we include all redundant checks, asymptotically there is no gain in the LP decoder threshold on the BSC under certain conditions on the base Tanner graph. First, we show that if the graph has bounded check-degree and satisfies a condition which we call asymptotic strength, then including high degree redundant checks in the LP does not significantly improve the threshold in the following sense: for each constant delta>0, there is a constant k>0 such that the threshold of the LP decoder containing all redundant checks of degree at most k improves by at most delta upon adding to the LP all redundant checks of degree larger than k. We conclude that if the graph satisfies a rigidity condition, then including all redundant checks does not improve the threshold of the base LP. We call the graph asymptotically strong if the LP decoder corrects a constant fraction of errors even if the LLRs of the correct variables are arbitrarily small. By building on the work of Feldman et al.(2007) and Viderman(2013), we show that asymptotic strength follows from sufficiently large expansion. We also give a geometric interpretation of asymptotic strength in terms pseudocodewords. We call the graph rigid if the minimum weight of a sum of check nodes involving a cycle tends to infinity as the block length tends to infinity. Under the assumptions that the graph girth is logarithmic and the minimum check degree is at least 3, rigidity is equivalent to the nondegeneracy property that adding at least logarithmically many checks does not give a constant weight check. We argue that nondegeneracy is a typical property of random check-regular graphs.
△ Less
Submitted 13 April, 2015; v1 submitted 27 November, 2014;
originally announced November 2014.
-
Weight distribution of cosets of small codes with good dual properties
Authors:
Louay Bazzi
Abstract:
The bilateral minimum distance of a binary linear code is the maximum $d$ such that all nonzero codewords have weights between $d$ and $n-d$. Let $Q\subset \{0,1\}^n$ be a binary linear code whose dual has bilateral minimum distance at least $d$, where $d$ is odd. Roughly speaking, we show that the average $L_\infty$-distance -- and consequently the $L_1$-distance -- between the weight distributio…
▽ More
The bilateral minimum distance of a binary linear code is the maximum $d$ such that all nonzero codewords have weights between $d$ and $n-d$. Let $Q\subset \{0,1\}^n$ be a binary linear code whose dual has bilateral minimum distance at least $d$, where $d$ is odd. Roughly speaking, we show that the average $L_\infty$-distance -- and consequently the $L_1$-distance -- between the weight distribution of a random cosets of $Q$ and the binomial distribution decays quickly as the bilateral minimum distance $d$ of the dual of $Q$ increases. For $d = Θ(1)$, it decays like $n^{-Θ(d)}$. On the other $d=Θ(n)$ extreme, it decays like and $e^{-Θ(d)}$. It follows that, almost all cosets of $Q$ have weight distributions very close to the to the binomial distribution. In particular, we establish the following bounds. If the dual of $Q$ has bilateral minimum distance at least $d=2t+1$, where $t\geq 1$ is an integer, then the average $L_\infty$-distance is at most $\min\{\left(e\ln{\frac{n}{2t}}\right)^{t}\left(\frac{2t}{n}\right)^{\frac{t}{2} }, \sqrt{2} e^{-\frac{t}{10}}\}$. For the average $L_1$-distance, we conclude the bound $\min\{(2t+1)\left(e\ln{\frac{n}{2t}}\right)^{t} \left(\frac{2t}{n}\right)^{\frac{t}{2}-1},\sqrt{2}(n+1)e^{-\frac{t}{10}}\}$, which gives nontrivial results for $t\geq 3$. We given applications to the weight distribution of cosets of extended Hadamard codes and extended dual BCH codes. Our argument is based on Fourier analysis, linear programming, and polynomial approximation techniques.
△ Less
Submitted 10 July, 2017; v1 submitted 25 August, 2014;
originally announced August 2014.
-
Linear Programming Decoding of Spatially Coupled Codes
Authors:
Louay Bazzi,
Badih Ghazi,
Rudiger Urbanke
Abstract:
For a given family of spatially coupled codes, we prove that the LP threshold on the BSC of the graph cover ensemble is the same as the LP threshold on the BSC of the derived spatially coupled ensemble. This result is in contrast with the fact that the BP threshold of the derived spatially coupled ensemble is believed to be larger than the BP threshold of the graph cover ensemble as noted by the w…
▽ More
For a given family of spatially coupled codes, we prove that the LP threshold on the BSC of the graph cover ensemble is the same as the LP threshold on the BSC of the derived spatially coupled ensemble. This result is in contrast with the fact that the BP threshold of the derived spatially coupled ensemble is believed to be larger than the BP threshold of the graph cover ensemble as noted by the work of Kudekar et al. (2011, 2012). To prove this, we establish some properties related to the dual witness for LP decoding which was introduced by Feldman et al. (2007) and simplified by Daskalakis et al. (2008). More precisely, we prove that the existence of a dual witness which was previously known to be sufficient for LP decoding success is also necessary and is equivalent to the existence of certain acyclic hyperflows. We also derive a sublinear (in the block length) upper bound on the weight of any edge in such hyperflows, both for regular LPDC codes and for spatially coupled codes and we prove that the bound is asymptotically tight for regular LDPC codes. Moreover, we show how to trade crossover probability for "LP excess" on all the variable nodes, for any binary linear code.
△ Less
Submitted 5 March, 2013; v1 submitted 27 January, 2013;
originally announced January 2013.
