This work deals with coding systems based on sparse graph codes. The key issue we address is the relationship between iterative (in particular belief propagation) and maximum a posteriori decoding. We show that between the two there is a fundamental connection, which is reminiscent of the Maxwell construction in thermodynamics. The main objects we consider are EXIT-like functions. EXIT functions were originally introduced as handy tools for the design of iterative coding systems. It gradually became clear that EXIT functions possess several fundamental properties. Many of these properties, however, apply only to the erasure case. This motivates us to introduce GEXIT functions that coincide with EXIT functions over the erasure channel. In many aspects, GEXIT functions over general memoryless output-symmetric channels play the same role as EXIT functions do over the erasure channel. In particular, GEXIT functions are characterized by the general area theorem. As a first consequence, we demonstrate that in order for the rate of an ensemble of codes to approach the capacity under belief propagation decoding, the GEXIT functions of the component codes have to be matched perfectly. This statement was previously known as the matching condition for the erasure case. We then use these GEXIT functions to show that in the limit of large blocklengths a fundamental connection appears between belief propagation and maximum a posteriori decoding. A decoding algorithm, which we call Maxwell decoder, provides an operational interpretation of this relationship for the erasure case. Both the algorithm and the analysis of the decoder are the translation of the Maxwell construction from statistical mechanics to the context of probabilistic decoding. We take the first steps to extend this construction to general memoryless output-symmetric channels. More exactly, a general upper bound on the maximum a posteriori threshold for sparse graph codes is given. It is conjectured that the fundamental connection between belief propagation and maximum a posteriori decoding carries over to the general case.
This work deals with coding systems based on sparse graph codes. The key issue we address is the relationship between iterative (in particular belief propagation) and maximum a posteriori decoding. We show that between the two there is a fundamental connection, which is reminiscent of the Maxwell construction in thermodynamics. The main objects we consider are EXIT-like functions. EXIT functions were originally introduced as handy tools for the design of iterative coding systems. It gradually became clear that EXIT functions possess several fundamental properties. Many of these properties, however, apply only to the erasure case. This motivates us to introduce GEXIT functions that coincide with EXIT functions over the erasure channel. In many aspects, GEXIT functions over general memoryless output-symmetric channels play the same role as EXIT functions do over the erasure channel. In particular, GEXIT functions are characterized by the general area theorem. As a first consequence, we demonstrate that in order for the rate of an ensemble of codes to approach the capacity under belief propagation decoding, the GEXIT functions of the component codes have to be matched perfectly. This statement was previously known as the matching condition for the erasure case. We then use these GEXIT functions to show that in the limit of large blocklengths a fundamental connection appears between belief propagation and maximum a posteriori decoding. A decoding algorithm, which we call Maxwell decoder, provides an operational interpretation of this relationship for the erasure case. Both the algorithm and the analysis of the decoder are the translation of the Maxwell construction from statistical mechanics to the context of probabilistic decoding. We take the first steps to extend this construction to general memoryless output-symmetric channels. More exactly, a general upper bound on the maximum a posteriori threshold for sparse graph codes is given. It is conjectured that the fundamental connection between belief propagation and maximum a posteriori decoding carries over to the general case.
There is a fundamental relationship between belief propagation (BP) and maximuma posterioridecoding. The case of transmission over the binary erasure channel was investigated in detail in a companion paper (C. MEacuteasson, A. Montanari, and R. Urbanke, "Maxwell's construction: The hidden bridge between iterative and maximum a posteriori decoding,"IEEE Transactions on Information Theory, submitted for publication). This paper investigates the extension to general memoryless channels (paying special attention to the binary case). An area theorem for transmission over general memoryless channels is introduced and some of its many consequences are discussed. We show that this area theorem gives rise to an upper bound on the maximuma posteriorithreshold for sparse graph codes. In situations where this bound is tight, the extrinsic soft bit estimates delivered by the BP decoder coincide with the correcta posterioriprobabilities above the maximuma posteriorithreshold. More generally, it is conjectured that the fundamental relationship between the maximuma posterioriprobability (MAP) and the BP decoder which was observed for transmission over the binary erasure channel carries over to the general case. We finally demonstrate that in order for the design rate of an ensemble to approach the capacity under BP decoding the component codes have to be perfectly matched, a statement which is well known for the special case of transmission over the binary erasure channel.
The extrinsic information transfer (EXIT) and generalized EXIT (GEXIT) area theorems provide fundamental conservation laws that connect the rate of classical codes to the performance of maximum-a-posteriori decoding on binary memoryless channels. In this work, we generalize these area theorems to quantum error correction. We first develop a coset-entropy formalism for nested linear codes, where the message specifies a coset and encoder randomization selects a codeword within that coset. The area under the resulting coset (G)EXIT function equals the normalized dimension of the quotient. For stabilizer codes, the measured syndrome restricts the Pauli error but does not always determine its logical effect, and the remaining uncertainty is an entropy over logical correction classes. We show that this entropy equals a classical coset entropy associated with the stabilizer group and the Pauli operators that commute with it. The resulting quantum GEXIT function has area equal to the number of encoded qubits per channel use for any stabilizer code, and this applies to the depolarizing channel through its joint four-valued Pauli-error distribution. For CSS codes on the quantum erasure channel and under independent bit- and phase-flip noise, the logical correction problem separates into two binary coset-decoding problems. On the quantum erasure channel, the duality between the stabilizer group and the Pauli operators that commute with it gives a folding identity at complementary erasure probabilities. Using this and the area theorem, we derive an upper bound on the rate of code families with vanishing logical error probability that matches the capacity of the quantum erasure channel. For the depolarizing channel, we also derive a rate-dependent converse, using a connection between correction-class entropy and coherent information to establish an inequality involving the complementary channel.
This thesis deals with the asymptotic analysis of coding systems based on sparse graph codes. The goal of this work is to analyze the decoder performance when transmitting over a general binary-input memoryless symmetric-output (BMS) channel. We consider the two most fundamental decoders, the optimal maximum a posteriori (MAP) decoder and the sub-optimal belief propagation (BP) decoder. The BP decoder has low-complexity and its performance analysis is, hence, of great interest. The MAP decoder, on the other hand, is computationally expensive. However, the MAP decoder analysis provides fundamental limits on the code performance. As a result, the MAP-decoding analysis is important in designing codes which achieve the ultimate Shannon limit. It would be fair to say that, over the binary erasure channel (BEC), the performance of the MAP and BP decoder has been thoroughly understood. However, much less is known in the case of transmission over general BMS channels. The combinatorial methods used for analyzing the case of BEC do not extend easily to the general case. The main goal of this thesis is to advance the analysis in the case of transmission over general BMS channels. To do this, we use the recent convergence of statistical physics and coding theory. Sparse graph codes can be mapped into appropriate statistical physics spin-glass models. This allows us to use sophisticated methods from rigorous statistical mechanics like the correlation inequalities, interpolation method and cluster expansions for the purpose of our analysis. One of the main results of this thesis is that in some regimes of noise, the BP decoder is optimal for a typical code in an ensemble of codes. This result is a pleasing extension of the same result for the case of BEC. An important consequence of our results is that the heuristic predictions of the replica and cavity methods of spin-glass theory are correct in the realm of sparse graph codes.
In the first part of this thesis we are interested in the asymptotic performance analysis of Non-Binary Low-Density Parity-Check (NBLDPC) codes over the Binary Erasure Channel (BEC) decoded via the suboptimal Belief Propagation (BP) decoder as well as the optimal Maximum-A-Posteriori (MAP) decoder. Previously, NBLDPC codes defined with respect to a finite field have been studied in the literature. Unfortunately, for these codes the asymptotic analysis of the BP decoder in terms of density evolution is cumbersome since the involved densities "live" in a high dimensional space. To alleviate this problem, we introduce ensembles defined with respect to the general linear group. For these ensembles the density evolution equations can be written in a compact form. We compute thresholds for different alphabet sizes for various NBLDPC ensembles. Surprisingly, the threshold is not a monotonic function of the alphabet size. We also conjecture the stability condition for NBLDPC ensembles defined with respect to finite fields over any binary memoryless symmetric channel. We then consider the performance of NBLDPC codes under MAP decoding when transmission takes place over the BEC. Towards this goal, we generalize the concepts of the peeling decoder and stopping sets to NBLDPC codes. Using these concepts, we give a combinatorial characterization of decoding failures for NBLDPC codes decoded via the BP decoder. We use the decoding failure criterion and the density evolution analysis to compute the asymptotic residual degree distribution for NBLDPC codes. The rate of this residual ensemble gives the conditional entropy of the NBLDPC code. To compute the rate of the residual ensemble we generalize the criterion of [1] to the non-binary setting, which, when satisfied, guarantees that almost every code in the ensemble has a rate equal to the design rate. We observe that the Maxwell construction of [1], relating the performance of MAP and BP decoding via the Extended BP GEXIT (EBP GEXIT [2]) function, holds in the setting of NBLDPC codes. The weight and stopping set distribution of a code are related to its performance under the MAP and the BP decoder respectively. We first concentrate on the weight distribution of regular NBLDPC ensembles. We derive the average weight distribution of regular NBLDPC ensembles. Using this, we derive the average of equivalent binary weight distribution of regular NBLDPC ensembles. We show that as the alphabet size becomes larger the equivalent binary weight distribution converges to the weight distribution of Gallager's random parity check ensemble. The average weight distribution of binary LDPC ensembles has been extensively studied in the literature. Much less is known about the probability distribution of this quantity. In particular, the question of the concentration of the weight distribution around the ensemble average has not been addressed. We compute the variance of the weight distribution of binary regular LDPC ensembles. Using this and the second moment method we obtain bounds on the probability that a randomly chosen code has its weight distribution close to the ensemble average. We show that a large fraction of the total number of codes have their weight distribution close to the average. We apply the same technique for the stopping set distribution of binary regular LDPC ensembles. Again we show that a large fraction of total number of codes have their stopping set distribution close to the ensemble average. In the last part of this thesis we address the question of the existence of the EBP GEXIT function. The EBP GEXIT function plays a fundamental role in the asymptotic analysis of sparse graph codes. For transmission over the BEC using binary LDPC codes, the analytic properties of the EBP GEXIT function are relatively simple and well understood. The case of general BMS channel is much harder and even the existence of the curve is not known in general. We introduce some tools from non-linear analysis which can be useful to prove the existence of EXIT like curves in some cases. The main tool is the Krasnoselskii-Rabinowitz (KR) bifurcation theorem.
The subject of this paper is transmission over a general class of binary-input memoryless symmetric channels using error correcting codes based on sparse graphs, namely, low-density generator-matrix and low-density parity-check codes. The optimal (or ideal) decoder based on the posterior measure over the code-bits and its relationship to the suboptimal belief propagation decoder are investigated. We consider the correlation (or covariance) between two code-bits, averaged over the noise realizations, as a function of the graph distance for the optimal decoder. Our main result is that this correlation decays exponentially fast for given low-density generator-matrix codes and a high enough noise parameter and also for given low-density parity-check codes and a low enough noise parameter. This has many consequences. Appropriate performance curves—called generalized extrinsic information transfer (GEXIT) functions—of the belief propagation and optimal decoders match in high/low noise regimes. This means that in high/low noise regimes the performance curves of the optimal decoder can be computed by density evolution. Another interpretation is that the replica predictions of spin-glass theory are exact. Our methods are rather general and use cluster expansions first developed in the context of mathematical statistical mechanics.
We consider communication over binary-input memoryless symmetric channels with low-density parity-check (LDPC) codes. The relationship between maximum a posteriori and belief propagation decoding is investigated using a set of correlation inequalities that first appeared in statistical mechanics of Gaussian spin glasses. We prove bounds on generalized extrinsic information transfer (EXIT) functions, that are believed to be tight, and discuss their relationship with the ones obtained by the interpolation method.