arXiv subjects

Eirik Rosnes

Publications and source records attributed to Eirik Rosnes.

At least 73 records · Page 4Linked to original sources

arXiv2018

An MDS-PIR Capacity-Achieving Protocol for Distributed Storage Using Non-MDS Linear Codes

We propose a private information retrieval (PIR) protocol for distributed storage systems with noncolluding nodes where data is stored using an arbitrary linear code. An expression for the PIR rate, i.e., the ratio of the amount of retrieved data per unit of downloaded data, is derived, and a necessary and a sufficient condition for codes to achieve the maximum distance separable (MDS) PIR capacity are given. The necessary condition is based on the generalized Hamming weights of the storage code, while the sufficient condition is based on code automorphisms. We show that cyclic codes and Reed-Muller codes satisfy the sufficient condition and are thus MDS-PIR capacity-achieving.

cs.IT↗

arXiv2018

Asymptotic Analysis and Spatial Coupling of Counter Braids

A counter braid (CB) is a novel counter architecture introduced by Lu et al. in 2007 for per-flow measurements on high-speed links which can be decoded with low complexity using message passing (MP). CBs achieve an asymptotic compression rate (under optimal decoding) that matches the entropy lower bound of the flow size distribution. In this paper, we apply the concept of spatial coupling to CBs to improve the performance of the original CBs and analyze the performance of the resulting spatially-coupled CBs (SC-CBs). We introduce an equivalent bipartite graph representation of CBs with identical iteration-by-iteration finite-length and asymptotic performance. Based on this equivalent representation, we then analyze the asymptotic performance of single-layer CBs and SC-CBs under the MP decoding algorithm proposed by Lu et al.. In particular, we derive the potential threshold of the uncoupled system and show that it is equal to the area threshold. We also derive the Maxwell decoder for CBs and prove that the potential threshold is an upper bound on the Maxwell decoding threshold. We show that the area under the extended MP extrinsic information transfer curve is equal to zero precisely at the area threshold. This, combined with the analysis of the Maxwell decoder and simulation results, leads us to the conjecture that the potential threshold is equal to the Maxwell decoding threshold and hence a lower bound on the maximum a posteriori decoding threshold. Interestingly, SC-CBs do not show the well-known phenomenon of threshold saturation of the MP decoding threshold to the potential threshold characteristic of spatially-coupled low-density parity-check codes and other coupled systems. However, SC-CBs yield better MP decoding thresholds than their uncoupled counterparts. Finally, we also consider SC-CBs as a compressed sensing scheme and show that low undersampling factors can be achieved.

cs.IT↗

arXiv2018

Lengthening and Extending Binary Private Information Retrieval Codes

It was recently shown by Fazeli et al. that the storage overhead of a traditional $t$-server private information retrieval (PIR) protocol can be significantly reduced using the concept of a $t$-server PIR code. In this work, we show that a family of $t$-server PIR codes (with increasing dimensions and blocklengths) can be constructed from an existing $t$-server PIR code through lengthening by a single information symbol and code extension by at most $\bigl\lceil t/2\bigr\rceil$ code symbols. Furthermore, by extending a code construction notion from Steiner systems by Fazeli et al., we obtain a specific family of $t$-server PIR codes. Based on a code construction technique that lengthens and extends a $t$-server PIR code simultaneously, a basic algorithm to find good (i.e., small blocklength) $t$-server PIR codes is proposed. For the special case of $t=5$, we find provably optimal PIR codes for code dimensions $k\leq 6$, while for all $7\leq k\leq 32$ we find codes of smaller blocklength than the best known codes from the literature. Furthermore, in the case of $t = 8$, we also find better codes for $k = 5, 6, 11, 12$. Numerical results show that most of the best found $5$-server PIR codes can be constructed from the proposed family of codes connected to Steiner systems.

cs.IT↗

arXiv2017

Block-Diagonal Coding for Distributed Computing With Straggling Servers

We consider the distributed computing problem of multiplying a set of vectors with a matrix. For this scenario, Li et al. recently presented a unified coding framework and showed a fundamental tradeoff between computational delay and communication load. This coding framework is based on maximum distance separable (MDS) codes of code length proportional to the number of rows of the matrix, which can be very large. We propose a block-diagonal coding scheme consisting of partitioning the matrix into submatrices and encoding each submatrix using a shorter MDS code. We show that the assignment of coded matrix rows to servers to minimize the communication load can be formulated as an integer program with a nonlinear cost function, and propose an algorithm to solve it. We further prove that, up to a level of partitioning, the proposed scheme does not incur any loss in terms of computational delay (as defined by Li et al.) and communication load compared to the scheme by Li et al.. We also show numerically that, when the decoding time is also taken into account, the proposed scheme significantly lowers the overall computational delay with respect to the scheme by Li et al.. For heavy partitioning, this is achieved at the expense of a slight increase in communication load.

cs.IT↗

arXiv2017

Private Information Retrieval in Distributed Storage Systems Using an Arbitrary Linear Code

We propose an information-theoretic private information retrieval (PIR) scheme for distributed storage systems where data is stored using a linear systematic code of rate $R > 1/2$. The proposed scheme generalizes the PIR scheme for data stored using maximum distance separable codes recently proposed by Tajeddine and El Rouayheb for the scenario of a single spy node. We further propose an algorithm to optimize the communication price of privacy (cPoP) using the structure of the underlying linear code. As an example, we apply the proposed algorithm to several distributed storage codes, showing that the cPoP can be significantly reduced by exploiting the structure of the distributed storage code.

cs.IT↗

arXiv2016

On Failing Sets of the Interval-Passing Algorithm for Compressed Sensing

In this work, we analyze the failing sets of the interval-passing algorithm (IPA) for compressed sensing. The IPA is an efficient iterative algorithm for reconstructing a k-sparse nonnegative n-dimensional real signal x from a small number of linear measurements y. In particular, we show that the IPA fails to recover x from y if and only if it fails to recover a corresponding binary vector of the same support, and also that only positions of nonzero values in the measurement matrix are of importance for success of recovery. Based on this observation, we introduce termatiko sets and show that the IPA fails to fully recover x if and only if the support of x contains a nonempty termatiko set, thus giving a complete (graph-theoretic) description of the failing sets of the IPA. Finally, we present an extensive numerical study showing that in many cases there exist termatiko sets of size strictly smaller than the stopping distance of the binary measurement matrix; even as low as half the stopping distance in some cases.

cs.IT↗

arXiv2016

Secure Repairable Fountain Codes

In this letter, we provide the construction of repairable fountain codes (RFCs) for distributed storage systems that are information-theoretically secure against an eavesdropper that has access to the data stored in a subset of the storage nodes and the data downloaded to repair an additional subset of storage nodes. The security is achieved by adding random symbols to the message, which is then encoded by the concatenation of a Gabidulin code and an RFC. We compare the achievable code rates of the proposed codes with those of secure minimum storage regenerating codes and secure locally repairable codes.

cs.IT↗

arXiv2015

Analysis of Spatially-Coupled Counter Braids

A counter braid (CB) is a novel counter architecture introduced by Lu et al. in 2007 for per-flow measurements on high-speed links. CBs achieve an asymptotic compression rate (under optimal decoding) that matches the entropy lower bound of the flow size distribution. Spatially-coupled CBs (SC-CBs) have recently been proposed. In this work, we further analyze single-layer CBs and SC-CBs using an equivalent bipartite graph representation of CBs. On this equivalent representation, we show that the potential and area thresholds are equal. We also show that the area under the extended belief propagation (BP) extrinsic information transfer curve (defined for the equivalent graph), computed for the expected residual CB graph when a peeling decoder equivalent to the BP decoder stops, is equal to zero precisely at the area threshold. This, combined with simulations and an asymptotic analysis of the Maxwell decoder, leads to the conjecture that the area threshold is in fact equal to the Maxwell decoding threshold and hence a lower bound on the maximum a posteriori (MAP) decoding threshold. Finally, we present some numerical results and give some insight into the apparent gap of the BP decoding threshold of SC-CBs to the conjectured lower bound on the MAP decoding threshold.

cs.IT↗

arXiv2015

On the Minimum Distance of Array-Based Spatially-Coupled Low-Density Parity-Check Codes

An array low-density parity-check (LDPC) code is a quasi-cyclic LDPC code specified by two integers $q$ and $m$, where $q$ is an odd prime and $m \leq q$. The exact minimum distance, for small $q$ and $m$, has been calculated, and tight upper bounds on it for $m \leq 7$ have been derived. In this work, we study the minimum distance of the spatially-coupled version of these codes. In particular, several tight upper bounds on the optimal minimum distance for coupling length at least two and $m=3,4,5$, that are independent of $q$ and that are valid for all values of $q \geq q_0$ where $q_0$ depends on $m$, are presented. Furthermore, we show by exhaustive search that by carefully selecting the edge spreading or unwrapping procedure, the minimum distance (when $q$ is not very large) can be significantly increased, especially for $m=5$.

cs.IT↗

arXiv2014

On the Minimum/Stopping Distance of Array Low-Density Parity-Check Codes

In this work, we study the minimum/stopping distance of array low-density parity-check (LDPC) codes. An array LDPC code is a quasi-cyclic LDPC code specified by two integers q and m, where q is an odd prime and m <= q. In the literature, the minimum/stopping distance of these codes (denoted by d(q,m) and h(q,m), respectively) has been thoroughly studied for m <= 5. Both exact results, for small values of q and m, and general (i.e., independent of q) bounds have been established. For m=6, the best known minimum distance upper bound, derived by Mittelholzer (IEEE Int. Symp. Inf. Theory, Jun./Jul. 2002), is d(q,6) <= 32. In this work, we derive an improved upper bound of d(q,6) <= 20 and a new upper bound d(q,7) <= 24 by using the concept of a template support matrix of a codeword/stopping set. The bounds are tight with high probability in the sense that we have not been able to find codewords of strictly lower weight for several values of q using a minimum distance probabilistic algorithm. Finally, we provide new specific minimum/stopping distance results for m <= 7 and low-to-moderate values of q <= 79.

cs.IT↗

arXiv2014

Minimum Pseudoweight Analysis of 3-Dimensional Turbo Codes

In this work, we consider pseudocodewords of (relaxed) linear programming (LP) decoding of 3-dimensional turbo codes (3D-TCs). We present a relaxed LP decoder for 3D-TCs, adapting the relaxed LP decoder for conventional turbo codes proposed by Feldman in his thesis. We show that the 3D-TC polytope is proper and $C$-symmetric, and make a connection to finite graph covers of the 3D-TC factor graph. This connection is used to show that the support set of any pseudocodeword is a stopping set of iterative decoding of 3D-TCs using maximum a posteriori constituent decoders on the binary erasure channel. Furthermore, we compute ensemble-average pseudoweight enumerators of 3D-TCs and perform a finite-length minimum pseudoweight analysis for small cover degrees. Also, an explicit description of the fundamental cone of the 3D-TC polytope is given. Finally, we present an extensive numerical study of small-to-medium block length 3D-TCs, which shows that 1) typically (i.e., in most cases) when the minimum distance $d_{\rm min}$ and/or the stopping distance $h_{\rm min}$ is high, the minimum pseudoweight (on the additive white Gaussian noise channel) is strictly smaller than both the $d_{\rm min}$ and the $h_{\rm min}$, and 2) the minimum pseudoweight grows with the block length, at least for small-to-medium block lengths.

cs.IT↗

arXiv2014

Using Short Synchronous WOM Codes to Make WOM Codes Decodable

In the framework of write-once memory (WOM) codes, it is important to distinguish between codes that can be decoded directly and those that require that the decoder knows the current generation to successfully decode the state of the memory. A widely used approach to construct WOM codes is to design first nondecodable codes that approach the boundaries of the capacity region, and then make them decodable by appending additional cells that store the current generation, at an expense of a rate loss. In this paper, we propose an alternative method to make nondecodable WOM codes decodable by appending cells that also store some additional data. The key idea is to append to the original (nondecodable) code a short synchronous WOM code and write generations of the original code and of the synchronous code simultaneously. We consider both the binary and the nonbinary case. Furthermore, we propose a construction of synchronous WOM codes, which are then used to make nondecodable codes decodable. For short-to-moderate block lengths, the proposed method significantly reduces the rate loss as compared to the standard method.

cs.IT↗

arXiv2014

Efficient Maximum-Likelihood Decoding of Linear Block Codes on Binary Memoryless Channels

In this work, we consider efficient maximum-likelihood decoding of linear block codes for small-to-moderate block lengths. The presented approach is a branch-and-bound algorithm using the cutting-plane approach of Zhang and Siegel (IEEE Trans. Inf. Theory, 2012) for obtaining lower bounds. We have compared our proposed algorithm to the state-of-the-art commercial integer program solver CPLEX, and for all considered codes our approach is faster for both low and high signal-to-noise ratios. For instance, for the benchmark (155,64) Tanner code our algorithm is more than 11 times as fast as CPLEX for an SNR of 1.0 dB on the additive white Gaussian noise channel. By a small modification, our algorithm can be used to calculate the minimum distance, which we have again verified to be much faster than using the CPLEX solver.

cs.IT↗

arXiv2014

Near-Field Passive RFID Communication: Channel Model and Code Design

This paper discusses a new channel model and code design for the reader-to-tag channel in near-field passive radio frequency identification (RFID) systems using inductive coupling as a power transfer mechanism. If the receiver resynchronizes its internal clock each time a bit is detected, the bit-shift channel used previously in the literature to model the reader-to-tag channel needs to be modified. In particular, we propose a discretized Gaussian shift channel as a new channel model in this scenario. We introduce the concept of quantifiable error avoidance, which is much simpler than error correction. The capacity is computed numerically, and we also design some new simple codes for error avoidance on this channel model based on insights gained from the capacity calculations. Finally, some simulation results are presented to compare the proposed codes to the Manchester code and two previously proposed codes for the bit-shift channel model.

cs.IT↗

arXiv2011

Further Results on Quadratic Permutation Polynomial-Based Interleavers for Turbo Codes

An interleaver is a critical component for the channel coding performance of turbo codes. Algebraic constructions are of particular interest because they admit analytical designs and simple, practical hardware implementation. Also, the recently proposed quadratic permutation polynomial (QPP) based interleavers by Sun and Takeshita (IEEE Trans. Inf. Theory, Jan. 2005) provide excellent performance for short-to-medium block lengths, and have been selected for the 3GPP LTE standard. In this work, we derive some upper bounds on the best achievable minimum distance dmin of QPP-based conventional binary turbo codes (with tailbiting termination, or dual termination when the interleaver length N is sufficiently large) that are tight for larger block sizes. In particular, we show that the minimum distance is at most 2(2^{ν+1}+9), independent of the interleaver length, when the QPP has a QPP inverse, where ν is the degree of the primitive feedback and monic feedforward polynomials. However, allowing the QPP to have a larger degree inverse may give strictly larger minimum distances (and lower multiplicities). In particular, we provide several QPPs with an inverse degree of at least three for some of the 3GPP LTE interleaver lengths giving a dmin with the 3GPP LTE constituent encoders which is strictly larger than 50. For instance, we have found a QPP for N=6016 which gives an estimated dmin of 57. Furthermore, we provide the exact minimum distance and the corresponding multiplicity for all 3GPP LTE turbo codes (with dual termination) which shows that the best minimum distance is 51. Finally, we compute the best achievable minimum distance with QPP interleavers for all 3GPP LTE interleaver lengths N <= 4096, and compare the minimum distance with the one we get when using the 3GPP LTE polynomials.

cs.IT↗

arXiv2011

Performance Analysis of 3-Dimensional Turbo Codes

In this work, we consider the minimum distance properties and convergence thresholds of 3-dimensional turbo codes (3D-TCs), recently introduced by Berrou et al.. Here, we consider binary 3D-TCs while the original work of Berrou et al. considered double-binary codes. In the first part of the paper, the minimum distance properties are analyzed from an ensemble perspective, both in the finite-length regime and in the asymptotic case of large block lengths. In particular, we analyze the asymptotic weight distribution of 3D-TCs and show numerically that their typical minimum distance dmin may, depending on the specific parameters, asymptotically grow linearly with the block length, i.e., the 3D-TC ensemble is asymptotically good for some parameters. In the second part of the paper, we derive some useful upper bounds on the dmin when using quadratic permutation polynomial (QPP) interleavers with a quadratic inverse. Furthermore, we give examples of interleaver lengths where an upper bound appears to be tight. The best codes (in terms of estimated dmin) obtained by randomly searching for good pairs of QPPs for use in the 3D-TC are compared to a probabilistic lower bound on the dmin when selecting codes from the 3D-TC ensemble uniformly at random. This comparison shows that the use of designed QPP interleavers can improve the dmin significantly. For instance, we have found a (6144,2040) 3D-TC with an estimated dmin of 147, while the probabilistic lower bound is 69. Higher rates are obtained by puncturing nonsystematic bits, and optimized periodic puncturing patterns for rates 1/2, 2/3, and 4/5 are found by computer search. Finally, we give iterative decoding thresholds, computed from an extrinsic information transfer chart analysis, and present simulation results on the additive white Gaussian noise channel to compare the error rate performance to that of conventional turbo codes.

cs.IT↗

arXiv2011

On the Analysis of Weighted Nonbinary Repeat Multiple-Accumulate Codes

In this paper, we consider weighted nonbinary repeat multiple-accumulate (WNRMA) code ensembles obtained from the serial concatenation of a nonbinary rate-1/n repeat code and the cascade of L>= 1 accumulators, where each encoder is followed by a nonbinary random weighter. The WNRMA codes are assumed to be iteratively decoded using the turbo principle with maximum a posteriori constituent decoders. We derive the exact weight enumerator of nonbinary accumulators and subsequently give the weight enumerators for WNRMA code ensembles. We formally prove that the symbol-wise minimum distance of WNRMA code ensembles asymptotically grows linearly with the block length when L >= 3 and n >= 2, and L=2 and n >= 3, for all powers of primes q >= 3 considered, where q is the field size. Thus, WNRMA code ensembles are asymptotically good for these parameters. We also give iterative decoding thresholds, computed by an extrinsic information transfer chart analysis, on the q-ary symmetric channel to show the convergence properties. Finally, we consider the binary image of WNRMA code ensembles and compare the asymptotic minimum distance growth rates with those of binary repeat multiple-accumulate code ensembles.

cs.IT↗

arXiv2009

Error Correcting Coding for a Non-symmetric Ternary Channel

Ternary channels can be used to model the behavior of some memory devices, where information is stored in three different levels. In this paper, error correcting coding for a ternary channel where some of the error transitions are not allowed, is considered. The resulting channel is non-symmetric, therefore classical linear codes are not optimal for this channel. We define the maximum-likelihood (ML) decoding rule for ternary codes over this channel and show that it is complex to compute, since it depends on the channel error probability. A simpler alternative decoding rule which depends only on code properties, called $\da$-decoding, is then proposed. It is shown that $\da$-decoding and ML decoding are equivalent, i.e., $\da$-decoding is optimal, under certain conditions. Assuming $\da$-decoding, we characterize the error correcting capabilities of ternary codes over the non-symmetric ternary channel. We also derive an upper bound and a constructive lower bound on the size of codes, given the code length and the minimum distance. The results arising from the constructive lower bound are then compared, for short sizes, to optimal codes (in terms of code size) found by a clique-based search. It is shown that the proposed construction method gives good codes, and that in some cases the codes are optimal.

cs.IT↗