We introduce the notion of the stopping redundancy hierarchy of a linear block code as a measure of the tradeoff between performance and complexity of iterative decoding for the binary erasure channel. We derive lower and upper bounds for the stopping redundancy hierarchy via Lovasz's local lemma (LLL) and Bonferroni-type inequalities, and specialize them for codes with cyclic parity-check matrices. Based on the observed properties of parity-check matrices with good stopping redundancy characteristics, we develop a novel decoding technique, termed automorphism group decoding, that combines iterative message passing and permutation decoding. We also present bounds on the smallest number of permutations of an automorphism group decoder needed to correct any set of erasures up to a prescribed size. Simulation results demonstrate that for a large number of algebraic codes, the performance of the new decoding method is close to that of maximum-likelihood (ML) decoding.
Abstract — We introduce the notion of the stopping redundancy hierarchy of a linear block code as a measure of the trade-off between performance and complexity of iterative decoding for the binary erasure channel. We derive lower and upper bounds for the stopping redundancy hierarchy via Lovász’s Local Lemma and Bonferroni-type inequalities, and specialize them for codes with cyclic parity-check matrices. Based on the observed prop-erties of parity-check matrices with good stopping redundancy characteristics, we develop a novel decoding technique, termed automorphism group decoding, that combines iterative message passing and permutation decoding. We also present bounds on the smallest number of permutations of an automorphism group decoder needed to correct any set of erasures up to a prescribed size. Simulation results demonstrate that for a large number of algebraic codes, the performance of the new decoding method is close to that of maximum likelihood decoding.
We extend the framework for studying the stopping redundancy of a linear block code by introducing and analyzing the stopping redundancy hierarchy.The stopping redundancy hierarchy of a code represents a measure of the trade-off between performance and complexity of iteratively decoding a code used over the binary erasure channel.It is defined as an ordered list of positive integers in which the ith entry, termed the i-th stopping redundancy, corresponds to the minimum number of rows in any parity-check matrix of the code that has stopping distance at least i.In particular, we derive lower and upper bounds for the i-th stopping redundancy of a code by using probabilistic methods and Bonferroni-type inequalities.Furthermore, we specialize the findings for cyclic codes, and show that parity-check matrices in cyclic form have some desirable redundancy properties.We also propose to investigate the influence of the generator codeword of the cyclic parity-check matrix on its stopping distance properties.
We investigate thestopping redundancy hierarchy oflinear blockcodesanditsconnection topermutation de- coding techniques. An element intheordered list ofstopping redundancy values represents thesmallest numberofpossibly linearly dependent rowsinanyparity-check matrix ofacode thatavoids stopping setsofup toa givensize. Redundant parity-check equations canbeshowntohavea similar effect ondecoding performance aspermuting thecoordinates ofthe received codeword according toaselected setofautomorphisms ofthecode.Basedonthisfinding we develop newdecoding strategies fordatatransmission overthebinary erasure channel thatcombine iterative message passing andpermutation decoding inordertoavoiderrors confined tostopping sets. We also introduce thenotion ofs-SADsets, containing thesmallest numberofautomorphisms ofacodewiththeproperty thatthey moveanysetofnotmorethanserasures intopositions that donotcorrespond tostopping setswithin ajudiciously chosen parity-check matrix.
It is now well known that the performance of a linear code
C under iterative decoding on a binary erasure channel (and other channels) is determined by the size of the smallest stopping set in the Tanner graph for
C. Several recent papers refer to this parameter as the \emph{stopping distance}
s of
C. This is somewhat of a misnomer since the size of the smallest stopping set in the Tanner graph for
C depends on the corresponding choice of a parity-check matrix. It is easy to see that
s≤d, where
d is the minimum Hamming distance of
C, and we show that it is always possible to choose a parity-check matrix for
C (with sufficiently many dependent rows) such that
s=d. We thus introduce a new parameter, termed the \emph{stopping redundancy} of
C, defined as the minimum number of rows in a parity-check matrix
H for
C such that the corresponding stopping distance
s(H) attains its largest possible value, namely
s(H)=d. We then derive general bounds on the stopping redundancy of linear codes. We also examine several simple ways of constructing codes from other codes, and study the effect of these constructions on the stopping redundancy. Specifically, for the family of binary Reed-Muller codes (of all orders), we prove that their stopping redundancy is at most a constant times their conventional redundancy. We show that the stopping redundancies of the binary and ternary extended Golay codes are at most 35 and 22, respectively. Finally, we provide upper and lower bounds on the stopping redundancy of MDS codes.
Stopping sets play a crucial role in failure events of iterative decoders over a binary erasure channel (BEC). The ℓth stopping redundancy is the minimum number of rows in the parity-check matrix of a code, which contains no stopping sets of size up to ℓ. In this paper, a notion of coverable stopping sets is defined. In order to achieve maximum-likelihood performance under iterative decoding over the BEC, the parity-check matrix should contain no coverable stopping sets of size ℓ, for 1 ≤ ℓ ≤ n-k, where n is the code length, k is the code dimension. By estimating the number of coverable stopping sets, we obtain upper bounds on the ℓth stopping redundancy, 1 ≤ ℓ ≤ n-k. The bounds are derived for both specific codes and code ensembles. In the range 1 ≤ ℓ ≤ d-1, for specific codes, the new bounds improve on the results in the literature. Numerical calculations are also presented.