Constraint Relaxation and Annealed Belief Propagation for Binary Networks

In this paper, we focus on the Generalized Belief Propagation (GBP) algorithm to solve trapping sets in Low-Density Parity-Check (LDPC) codes. Trapping sets are topological structures in Tanner graphs of LDPC codes that are not correctly decoded by Belief Propagation (BP), leading to exhibit an error-floor in the Bit-Error Rate (BER). Stemming from statistical physics of spin glasses, GBP consists in passing messages between clusters of Tanner graph nodes in another graph called the region-graph. Here, we introduce a specific clustering of nodes, based on a novel local loopfree principle, that breaks trapping sets such that the resulting region-graph is locally loopfree. We then construct a hybrid decoder made of BP and GBP that proves to be a powerful decoder as it clearly improves the BER and defeats the error-floor.

The cache-enabled D2D network supports direct interconnection among users and can help to offload the data traffic through user caching. In this paper, we consider a reasonable caching strategy design in the cache-enabled D2D network to improve its performance. First we formulate an optimization problem aiming to minimize the average file download time of the users. Next we transform the original problem to a max-product form that can be distributed solved. Then we introduce the factor graph and the conventional belief propagation (BP) algorithm. Finally, we show how we design a BP algorithm with Randomized Gaussian Message Passing scheme and Weighting mechanism (BP-RW) to improve its convergence and performance. The distributed BP-RW algorithm can assign almost all computations to each user in the network, which will help to reduce the heavy computational overhead of the central controller. Meanwhile, simulation results show that the BP-RW algorithm performs better than the conventional BP algorithm, and performs very closely to the centralized greedy algorithm with much lower computational complexity.

The major obstacle to shallow water acoustic communication is the interference of multipath signal results from surface and bottom reflections. Channel coding is indispensable in practical system to accommodate these adverse channel transmission characteristics due to the coding gain. In this paper, a channel model including time-varying fading, multipath and additive noise for the shallow water acoustic channels is built. Based on the channel model, the Probability Density Function (PDF) of initial decoding messages with Belief Propagation (BP) algorithm of LowDensity Parity-Check (LDPC) codes is deduced. The performance of the LDPC codes with BP algorithm is simulated, and the effects of multipath, fading velocity and channel interleaver on the decoding performance are studied. As the result, the following conclusions can be given: The performance of LDPC codes is excellent (Bit Error Rate (BER) is less than 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-4</sup> ) over three-multipath shallow water acoustic channels, with 3~5 iterative decoding times and about 1000 bits code lengths; The BER of LDPC codes degrades with the increasing of the number of multipath; The BER of LDPC codes degrades slowly with the decreasing of the fading velocity, which shows the proposed channel code is not sensitive to the fading velocity of the channel.

In this paper, we develop a Belief Propagation (BP) algorithm for similarity computation to improve the recommendation accuracy of the neighborhood method, which is one of the most popular Collaborative Filtering (CF) recommendation algorithms. We formulate a probabilistic inference problem as to compute the marginal posterior distributions of similarity variables from their joint posterior distribution given the observed ratings. However, direct computation is prohibitive in large-scale recommender systems. Therefore, we introduce an appropriate chosen factor graph to express the factorization of the joint distribution function, and utilize the BP algorithm that operates in the factor graph to exploit the factorization for efficient inference. In addition, since the high degree at the factor node incurs an exponential increase in computational complexity, we also propose a complexity-reduction technique. The overall complexity of the proposed BP algorithm on a factor graph is linear in the number of variables, which ensures scalability. Finally, through experiments on the MovieLens dataset, we show the superior prediction accuracy of the proposed BP-based similarity computation algorithm for recommendation.

We proposed two improved decoding algorithms for Low-Density Parity-Check (LDPC) codes based on the Belief-Propagation (BP) algorithm combined with Genetic Algorithm (GA). After giving a genetic interpretation of Tanner graph, GA is adopted to efficiently use the information passing from the variable nodes to the check nodes. Simulation results assert the superiority of our proposed algorithms over the BP algorithm both in BER (Bit Error Rate) and FER (Frame Error Rate). At last, optimization of the key parameter of the developed algorithms is given.

In this letter, a new early termination method is proposed for Luby transform (LT) belief propagation (BP) decoder. The proposed method is decided whether BP decoder output converges to original data bits by observing only sign alterations of log-likelihood ratio (LLR) messages in BP decoder structure. Simulation results and complexity analyzes show that proposed method has low computational complexity and small average iteration amounts compared to conventional early termination methods in literature. In addition to this, the proposed method doesn't cause any performance degradation in BP decoder. The method can be easily applied to code families which can be decoded by BP algorithm such as low density parity check (LDPC) codes, polar codes and Raptor codes.

In this paper, we investigate distributed inference schemes, over binary-valued Markov random fields, which are realized by the belief propagation (BP) algorithm. We first show that a decision variable obtained by the BP algorithm in a network of distributed agents can be approximated by a linear fusion of all the local log-likelihood ratios. The proposed approach clarifies how the BP algorithm works, simplifies the statistical analysis of its behavior, and enables us to develop a performance optimization framework for the BP-based distributed inference systems. Next, we propose a blind learning-adaptation scheme to optimize the system performance when there is no information available a priori describing the statistical behavior of the wireless environment concerned. In addition, we propose a blind threshold adaptation method to guarantee a certain performance level in a BP-based distributed detection system. To clarify the points discussed, we design a novel linear-BP-based distributed spectrum sensing scheme for cognitive radio networks and illustrate the performance improvement obtained, over an existing BP-based detection method, via computer simulations.

Severe intersymbol interference (ISI) is one of the main obstacles for reliable signal reception in ATSC DTV systems. Decision feedback equalizers (DFEs) are commonly used to suppress the ISI. However, DFEs may suffer from error propagation due to incorrect symbol decisions from the symbol slicer. This phenomenon deteriorates the performance even more when the post-cursor ISI is strong. In order to reduce error propagation, we present a novel hybrid equalization scheme for ATSC channels. The proposed scheme consists of an adaptive decision feedback sparsening filter (DFSF), and an iterative maximum a posteriori (MAP) equalizer based on the belief propagation (BP) algorithm. In the first stage, instead of removing all the ISI from post cursors, the DFSF employs a modified feedback filter which leaves the strongest post-cursor ISI taps uncorrected. As a result, a long ISI channel is equalized to a sparse channel having only a small number of nonzero taps. In the second stage, a belief propagation algorithm is applied to mitigate the residual ISI. Since the channel is typically time-varying and suffers from Doppler fading, the DFSF is adapted using the least mean square (LMS) algorithm, such that the amplitude and the locations of the nonzero taps of the equalized sparse channel appear to be fixed. As such, the channel appears to be static during the second stage of equalization which consists of the BP detector. Simulation results demonstrate that the proposed scheme outperforms the traditional DFE in symbol error rate, under both static channels and dynamic ATSC channels.

In high density patterned media storage system, the “multiple islands per read head” model is taken into account for inter-track interference (ITI) in bit-patterned media (BPM). It has been proposed that low-density parity-check (LDPC) codes could improve the symbol error rate (SER) performance significantly when only considering AWGN noise alone over the storage channel. LDPC decoding algorithm could be applied in LDPC joint decoder of patterned media storage system. In this paper, we proposed a decoding algorithm of serial belief propagation (SBP) with a stop updating criterion for LDPC decoder. Simulation results show that the proposed decoding algorithm could reduce the times of generating and propagating the message in decoding with negligible error rate performance loss over AWGN channels when appropriately choosing the parameter that determines which variable node has been converged. By taking advantage of extrinsic information transfer (EXIT) charts, it is demonstrated that the proposed decoding algorithm could maintain superior SER performance and convergence rate compared with the traditional SBP decoding algorithm. To further explore the performance of the proposed decoding algorithm in “multiple islands per read head” model, we performed a large number of simulations over this channel model with different distribution percentages between the AWGN and the island position jitter noise. Simulation results show that the LDPC codes could improve the SER performance when considering jitter noise while the proposed decoding algorithm behaves excellently over the patterned media storage channel and over the AWGN channel.

For industrial application of load monitoring techniques, it is important to establish high-performance state estimators of low-cost and low frequency smart meters (SM) and sensors in a power system, which can run under resource-constrained computing units. Because household electronic appliances often tap power from fixed sockets, a finite state table for the corresponding sensors is suitable and convenient. However, SM in the main line may have an enormous state table. In this study, we propose a belief propagation (BP) algorithm to calculate the power consumption of electronic appliances in a semi-intrusive load monitoring (SILM) system whose SM and sensors have state tables with sizes varying largely. The novelty of the proposed method lies in a continuous approximation to a large state table and a switching scheme between discrete and continuous parts of the SILM system. With datasets from numerical simulations and a real-world experimental SILM system in a set of high-density school buildings within a secondary distribution network, the proposed BP algorithm is compared with relevant state-of-the-art algorithms. The results show that the proposed algorithm achieves a percentage of error (8&#x0025;), which outperforms the percentage achieved by the other methods, a linear state estimation of 99&#x0025;, a hidden Markov model of 21&#x0025;, and a full-discrete BP algorithm of 11&#x0025;. In addition, the complexity of the proposed algorithm is the least of all methods, and the proposed algorithm can run by SoC on concentrators.

In this paper, we use the minimum mean square error (MMSE) criterion to design the scaled and offset belief propagation (BP)-based decoding algorithms for short low-density parity-check (LDPC) codes on the fast Rayleigh fading channel. Based on the MMSE criterion, the theoretical formulas and numerical calculations on the optimum factors for these two improved BP-based algorithms are provided. The simulation results for several short (Chen et al., 2005 and Gallager, 1962) regular LDPC codes on the fast Rayleigh fading channel demonstrate that the scaled and offset BP-based algorithms with our proposed factors can achieve the performance better than that of the BP algorithm.

Low density parity-check (LDPC) codes are a class of linear block codes that are decoded by running belief propagation (BP) algorithm or log-likelihood ratio belief propagation (LLR-BP) over the factor graph of the code. One of the disadvantages of LDPC codes is the onset of an error floor at high values of signal to noise ratio caused by trapping sets. In this paper, we propose a two stage decoder to deal with different types of trapping sets. Oscillating trapping sets are taken care by the first stage of the decoder and the elementary trapping sets are handled by the second stage of the decoder. Simulation results on the regular PEG (504,252,3,6) code and the irregular PEG (1024,518,15,8) code shows that the proposed two stage decoder performs significantly better than the standard decoder.

A BP-based algorithm with 2-dimensional classified normalized correction is developed to reduce the complexity and improve the performance of decoding algorithm for the low density parity check (LDPC) codes. The algorithm first utilizes classification according to the absolute values of incoming messages in check nodes. Then it uses 2-dimensional normalization to correct the minimum and sub-minimum values. The 2-dimensional normalized factors can be calculated respectively by using probability and statistic theory in the initialization step. Simulation results illustrate that the proposed algorithm achieves better performance in bit error ratio (BER) and average iteration number than normalized BP-based algorithm in the high signal noise ratio (SNR) region, i.e., it can achieve 0.4 dB SNR gain and reduce 20% number of iterations at BER= 5 10  , whose complexity is also much less than that of belief propagation (BP) algorithm. It is concluded that the proposed algorithm offers better tradeoff between performance and complexity.  Index Terms—channel coding theory, density evolution, Iterative decoding algorithm

This paper presents a general decoder architecture for 5G LDPC and polar codes. To achieve the hardware reuse maximization, both LDPC and polar codes are decoded by belief propagation (BP) algorithm on sparse bipartite graphs. However, by counting the number of short cycles on the polar bipartite graphs, we observe that numerous short cycles cause the poor performance than conventional decoding algorithms, like successive cancellation (SC) and BP decoding. To reduce the impact of short cycles, a LDPC-like BP list (BPL) decoding algorithm is proposed for polar codes, which is applied on L different polar bipartite graphs constructed by different permuted polar factor graphs. Finally, the simulation shows that the proposed general decoder for polar codes can obtain beyond SC and BP error performance with lower decoding latency and computational complexity, and even near successive cancellation list (SCL) performance in high signal-noise ratios (SNRs).

The neural networks offer iteration capability for low-density parity-check (LDPC) decoding with superior performance at transmission. However, to cope with increasing code length and rate, the complexity of the neural network increases significantly. This is due to the large amount of feature extraction required to maintain the error correction capability. Based on this gap, we design a new iterative LDPC decoding technique named graph model neural network-belief propagation (GMNN-BP). GMNN-BP uses graph models as a link between deep learning and belief propagation (BP) algorithms, combining the advantages of both. Compared to traditional fully connected neural network decoders, the GMNN-BP decoding has the substantial benefit of avoiding learning and judging codeword categories directly from a large amount of data and requiring less training data as well. The proposed algorithm is verified by simulation and experiment and is tested by using IEEE 802.3ca standard LDPC code word. The results show that the GMNN-BP decoding algorithm is superior to the BP-based iterative decoding method under the same number of iterations, and the maximum gain can reach 1.9dB. When achieving the same performance, the GMNN-BP decoding algorithm only requires half the number of iterations of other algorithms.