We study irregular repetition slotted ALOHA (IRSA) with multi-antenna reception over Rayleigh block fading channels. An exact closed-form expression for the average decoding error probability ¯ϵm,L is derived for collision sizes m = 1 and m = 2 by applying the inclusion–exclusion principle, which generalizes known single-antenna results and remains valid for any finite number of antennas. Using this result, we develop a density-evolution-based analysis of multi-antenna IR-SA systems and characterize belief-propagation (BP) thresholds. Numerical results for the corresponding maximum a posteriori (MAP) decoding thresholds and converse bounds are also presented, demonstrating threshold saturation with spatial coupling.

Existing tile size selection approaches are tightly coupled with compiler transformation pipelines, often leading to inaccurate modeling of cache behavior and limited effectiveness for non-rectangular tile shapes. This article presents TileMind , a decoupled analytical model that combines compile-time and runtime information for tile size selection in affine programs. It introduces a transformation-aware pre-tiling step that enables the decoupled selector to remain consistent with compiler transformations while extracting compile-time metadata. The extracted metadata is then combined with profiled runtime characteristics to construct a richer yet tractable feasible domain, within which a nonlinear objective for tile size selection is formulated. This objective is subsequently transformed into a binary product linearization problem, with its nonlinear constraints also linearized for efficient optimization. Finally, an intra-tile optimization aligns computation with data layout to enhance data reuse within tiles. Across two multi-core Intel CPUs, TileMind achieves 1.49× (sequential) and 1.33× (parallel) mean speedups on twenty PolyBench kernels, and 2.08–3.54× speedups on three deep learning workloads over the state-of-the-art analytical model Pluto-tss . Compared with TVM’s latest autotuner MetaSchedule, TileMind delivers 1.35–1.46× mean speedups while reducing tuning overhead by 2–4 orders of magnitude. While demonstrating effectiveness on selecting tile sizes for non-rectangular tile shapes and compatibility with PPCG, Pluto, and TVM, we further provide proof-of-concept results on GPUs, illustrating the potential portability of TileMind across architectures.

Resistive Random-Access Memory (ReRAM)—a core component of nanoscale mixed-signal storage circuits—faces a unique data-dependent challenge, sneak path (SP) interference, which severely limits its use in high-density integrated systems. With the rapid advancement of high-density ReRAM crossbar architectures, independently designed constrained coding or error-correcting coding (ECC) often introduces excessive hardware redundancy. It fails to synergistically optimize SP suppression and channel noise correction, reducing the integration and energy efficiency of ReRAM-based circuits. To address this, this paper proposes a unified multi-module collaborative storage error-control framework, achieving “suppression-correction” dual optimization for SP interference in one system. Constrained codes built on the maximum mutual information principle significantly lower SP occurrence probability while preserving channel capacity. An optimized adaptive belief propagation (BP) detection algorithm enables precise estimation of ReRAM array-stored data, and polar codes efficiently correct residual errors. Simulation results show the unified framework greatly improves array data discrimination accuracy. This work not only offers a new paradigm for balancing reliability and efficiency in practical ReRAM storage circuit design but also connects information theory with nanoscale mixed-signal circuit implementation.

An analytical framework integrating performance characterization and coding theory is proposed to mitigate sneak path (SP) interference in resistive random-access memory (ReRAM) crossbar arrays. The core innovation is identified in the mathematical decomposition of ReRAM’s non-ergodic data-dependent channel into multiple stationary memoryless subchannels. Through information-theoretic analysis, an approximate finite-length characterization of the theoretical lower bound for decoding word error probability (WEP) is established. This is achieved by systematically analyzing the SP occurrence rate in constrained array geometries combined with comprehensive evaluation of both mutual information and dispersion metrics across the decomposed channel components. Building upon this decomposition paradigm, a systematic code construction methodology is developed using density evolution principles for sparse-graph code design. The designed codes not only exhibit capacity-approaching decoding thresholds but also yield word error rate simulation results that are close to the derived WEP bound under practical crossbar configurations.

Low Earth orbit (LEO) satellite networks are expected to support heterogeneous services, including enhanced mobile broadband (eMBB) communications, massive machine-type communications (mMTC), and ultra-reliable low-latency communications (URLLC). However, existing coexistence schemes, such as puncturing and superposition, struggle to achieve an effective trade-off among reliability, latency, and spectral efficiency due to their limited degrees of freedom (DoF). To address this challenge, we propose a novel rate-splitting grant-free (RS-GF) transmission scheme that integrates rate-splitting multiple access (RSMA) with grant-free random access (GF-RA) to efficiently support heterogeneous quality of service (QoS) requirements. The high-rate eMBB user employs single-layer rate splitting (RS) over the entire slot, while short-packet Internet-of-Things (IoT) devices associated with URLLC and mMTC adopt GF-RA via single mini-slot transmissions. Building on this RS-GF framework, we analyze the outage performance of the proposed scheme. Specifically, we derive the average packet error probability (PEP) of IoT devices in the finite blocklength (FBL) regime and analyze the eMBB user’s outage probability under imperfect successive interference cancellation (SIC) and mini-slot collisions. On this basis, we present simplified analytical solutions for sparse and dense IoT deployment scenarios, and Monte Carlo simulations validate our analytical derivations. Simulation results demonstrate that the proposed RS-GF scheme outperforms state-of-the-art solutions for service coexistence in LEO satellite networks.

Transformer models face scalability challenges in causal language modeling (CLM) due to inefficient memory allocation for growing keyvalue (KV) caches, which strains compute and storage resources.Existing methods like Grouped Query Attention (GQA) and tokenlevel KV optimization improve efficiency but rely on rigid resource allocation, often discarding "low-priority" tokens or statically grouping them, failing to address the dynamic spectrum of token importance.We propose mixSGA, a novel mixture-of-expert (MoE) approach that dynamically optimizes token-wise computation and memory allocation.Unlike prior approaches, mixSGA retains all tokens while adaptively routing them to specialized experts with varying KV group sizes, balancing granularity and efficiency.Our key novelties include: (1) a token-wise expert-choice routing mechanism guided by learned importance scores, enabling proportional resource allocation without token discard; (2) weight-sharing across grouped attention projections to minimize parameter overhead; and (3) an auxiliary loss to ensure one-hot routing decisions for training-inference consistency in CLMs.Extensive evaluations across Llama3, TinyLlama, OPT, and Gemma2 model families show mixSGA's superiority over static baselines.On instruction-following and continued pretraining tasks, mixSGA achieves higher ROUGE-L and lower perplexity under the same KV budgets.

Multilingual Automatic Speech Recognition (ASR) systems face challenges in efficiently integrating new languages while maintaining scalability and performance. This paper introduces an Incremental Tucker-based Low-Rank Adaptation (LoRA) framework designed to address these limitations. By leveraging Tucker decomposition to compress multilingual LoRA parameters and introducing an efficient incremental learning mechanism, the framework enables seamless addition of new languages without re-computing the full decomposition. Experiments on the Mozilla Common Voice dataset demonstrate that the proposed approach reduces parameter requirements by up to 80% compared to traditional LoRA-based methods, while achieving competitive Word Error Rates (WER) across diverse languages. This framework provides a scalable and resource-efficient solution for modern multilingual ASR systems, offering significant advantages in memory efficiency and adaptability to dynamic language expansion.

The sneak path (SP) issue presents a substantial challenge for resistive random-access memory (ReRAM), significantly affecting data storage reliability. The SP rate, which represents the proportion of memory cells impacted by SPs, is a crucial parameter influencing the probability of data detection errors. In this paper, we concentrate on analyzing the probability distribution of the SP rate in ReRAM arrays that incorporate imperfect selectors. Our research indicates that when ReRAM stores data following an independent and identically distributed (i.i.d.) Bernoulli distribution with parameter$q$, and the array size is large, the SP rate approximates a Gaussian distribution. The mean and variance of this distribution can be explicitly derived as functions of the number of selector failures, parameter$q$, and the array size.

Resistive random-access memory (ReRAM) is a promising non-volatile memory technology. However, its crossbar array structure leads to a severe problem known as sneak path interference (SPI), which is correlated and data-dependent. From an information-theoretic perspective, memory systems like ReRAM can be considered as special types of communication channels. Inspired by deep learning applications in communication systems, the detection of ReRAM channels with SPI was formulated as a learning problem recently, and a multi-layer perceptron (MLP) network was employed to mitigate SPI. However, it requires a large amount of training data to achieve satisfactory performance. In this paper, we first propose a bidirectional long short-term memory (BiLSTM) based detector for ReRAM to exploit the correlation between memory cells introduced by SPI. Moreover, a few-shot learning algorithm based on meta-transfer learning (MTL) is proposed to further improve the generalization ability of the detector. The bit error rate (BER) bound and generalization bound are also derived to verify the effectiveness of our proposed schemes. Simulation results demonstrate that the BiLSTM-based detector with MTL can dramatically reduce the required training samples by four to five orders of magnitude while improving the BER performance compared to the existing MLP-based detection scheme.

Iterative decoding algorithms based on the alternating direction method of multipliers (ADMM) decoding of low density parity check (LDPC) codes has emerged as an alternating decoding method and bringed a boom of research on drawing upon mathematical optimization to LDPC decoding. Improving error-correcting performance is a key issue to enhance the superiority of ADMM decoding. In this letter, we investigate an efficient ADMM-based iterative decoder for binary LDPC codes. First, we build an mathematical programming equivalence of the maximum likelihood (ML) decoding problem by transforming parity-check constraints to multiple equivalent linear constraints and eliminating check-polytope projection (CPP). Then, an iterative algorithm based on ADMM technique is developed to solve this free-of-CPP (FCPP) equivalence and each ADMM update can be computed efficiently. Moreover, the proposed ADMM-FCPP decoding algorithm is analyzed to display a linear complexity to the length of the LDPC code at each iteration. Finally, simulation results demonstrate the superiority of the proposed decoder in error-correcting performance compared with the state-of-the-art ADMM-based decoders.

An enhanced on-off division multiple access (ODMA) transmission scheme is introduced for unsourced multiple access networks. Building upon the foundational ODMA transmission scheme, we have implemented further refinements to the original joint on-off pattern and data detection algorithm. Specifically, we propose a pattern collision resolution technique that can blindly recognize the collision degree of each on-off pattern, and then iteratively recover the data of collided users over a joint factor graph. Furthermore, we introduce a finite-length performance analysis for on-off pattern detection and iterative multi-user decoding. Through this analysis, we derive numerous numerical results, revealing the impact of various parameters on the performance of collision degree detection and multi-user decoding, respectively. By summarizing the rules observed from these numerical results, we formulate design strategies for these parameters, aiming to optimize the overall performance of our scheme. The inherent super sparse property of ODMA ensures that our scheme maintains low decoding complexity. Numerical results demonstrate that, with the implementation of our pattern collision resolution method and meticulous parameter design, the proposed scheme achieves a gap of less than 1.2 dB compared to the random coding bound for up to 300 active users. This performance surpasses state-of-the-art schemes across a broad range of user numbers.

The achievable rate of code over resistive random-access memory (ReRAM) channel with finite selector failures was published in our recent work. The rate was derived under the assumption of independent and identically distributed (i.i.d.) input. In this work, focusing on the ReRAM channel with a single selector failure in the memory array, we derive an upper bound on achievable rate under arbitrary input distribution. This upper bound is within 0.02 bits from the achievable rate of i.i.d. input, indicating that i.i.d. is very close to optimal for large memory arrays. Moreover, we analyze the achievable rate of random code over ReRAM channel with suboptimal decodings where the decoder ignores the channel correlation. Our result indicates that in this case the achievable rate is limited by the capacity of a memoryless channel. We reveal both weak and strong asymptotic properties of ReRAM channel to prove this. The proof can be directly extended to the case of ReRAM with an arbitrary number of selector failures in the memory array.

The achievable rate of code over resistive random-access memory (ReRAM) channel with finite selector failures was published in our recent work. The rate was derived under the assumption of independent and identically distributed (i.i.d.) input. In this work, focusing on the ReRAM channel in the case of single selector failure, we derive an upper bound on achievable rate under arbitrary input distribution. This upper bound is within 0.02 bits from the achievable rate of i.i.d. input, indicating that i.i.d. is very close to optimal for large memory arrays.

An enhanced on-off division multiple access (ODMA) transmission scheme is proposed for unsourced multiple access network. The message of each active user is divided into two parts, where the first part is used to determine an on-off pattern, and the second part is encoded and transmitted in a time-hopping manner according to an on-off pattern. Leveraging the super sparse property of ODMA, the users' on-off pattern and pattern collisions are blindly detected based on the received signal without the help of pilot. Moreover, the on-off pattern detection, data decoding and collision recovery are performed iteratively over one sparse graph to enhance the overall system reliablity. We propose a finite-length performance analysis to the on-off pattern detection and iterative multi-user decoding, based on which both the user access sparsity, and channel code are optimized. Numerical result shows that with a rate 1/ 3 low-density parity-check code over G F (26), the gap between the proposed scheme and the random coding bound is less than 1.2 dB for up to 300 active users.

Throughput analysis for successive interference cancellation-based two-device slotted ALOHA with feedback is studied over Nakagami-m fading channels. Explicit expressions for the state transition probabilities are derived for a Markov process, thus facilitating the computation of the throughput. Through optimization of the transmission probability, it is shown that the maximum throughput is achieved for a given code rate.

Reducing floating-point (FP) precision is used to trade the quality degradation of a numerical program's output for performance, but this optimization coincides with type casting, whose overhead is undisclosed until a mixed-precision code version is generated. This uncertainty enforces the decoupled implementation of mixed-precision code generation and autotuning in prior work. In this paper, we present a holistic approach called PrecTuner that consolidates the mixed-precision code generator and the autotuner by defining one parameter. This parameter is first initialized by some automatically sampled values and used to generate several code variants, with various loop transformations also taken into account. The generated code variants are next profiled to solve a performance model formulated using the aforementioned parameter, possibly under a pre-defined quality degradation budget. The best-performing value of the defined parameter is finally predicted without evaluating all code variants. Experimental results of the PolyBench benchmarks on CPU demonstrate that PrecTuner outperforms LuIs by 3.28× while achieving smaller errors, and we also validate its effectiveness in optimizing a real-life large-scale application. In addition, PrecTuner also obtains a mean speedup of 1.81× and 1.52×-1.73× over Pluto on single- and multi-core CPU, respectively, and 1.71× over PPCG on GPU.

Massive machine-type communications (mMTCs) features a massive number of low-cost user equipment (UE) with sparse activity. Tailor-made for these features, grant-free random access (GF-RA) serves as an efficient access solution for massive machine-type communication (mMTC). However, most existing GF-RA schemes rely on strict synchronization, which incurs excessive coordination burden for the low-cost UEs. In this work, we propose a receiver design for asynchronous GF-RA, and address the joint user-activity detection (UAD) and channel estimation (CE) problem in the presence of asynchronization-induced intersymbol interference. Specifically, the delay profile is exploited at the receiver to distinguish different UEs. However, a sample correlation problem in this receiver design impedes the factorization of the joint likelihood function, which complicates the UAD and CE problem. To address this correlation problem, we design a partially uni-directional (PUD) factor graph representation for the joint likelihood function. Building on this PUD factor graph, we further propose a PUD message passing-based sparse Bayesian learning (SBL) algorithm for asynchronous UAD and CE (PUDMP-SBL-aUADCE). Our theoretical analysis shows that the PUDMP-SBL-aUADCE algorithm exhibits higher signal-to-interference-and-noise ratio (SINR) in the asynchronous case than in the synchronous case, i.e., the proposed receiver design can exploit asynchronization to suppress multiuser interference. In addition, considering potential timing error from the low-cost UEs, we investigate the impacts of imperfect delay profile, and reveal the advantages of adopting the SBL method in this case. Finally, extensive simulation results are provided to demonstrate the performance of the PUDMP-SBL-aUADCE algorithm.

A signal scrambling based joint blind channel estimation, activity detection, and data decoding (SS-JCAD) scheme is proposed for coded massive random access. This signal scrambling technique imposes symbol-wise phase rotation to each user’s modulated data, and the scrambling pattern serves as a user-specific signature which is free from any bandwidth expansion or pilot signaling overhead. Building on this scrambling signature, we further propose a simple yet efficient receiver design, which integrates the blind channel state information (CSI) estimation module with the forward error correction (FEC) decoder. Specifically, according to the scrambling signature, a user-specific posterior probability density function of the CSI is derived, based on which both the CSI and activity of each user can be blindly detected using a low-complexity single-user maximum a posteriori estimation. Given the estimated CSI asa prioriinformation, a joint CSI (including user activity) estimation and data decoding algorithm is proposed, where the soft information is iteratively updated between the FEC decoder and the CSI estimation module to refine the detection reliability. Simulation shows that for massive random access systems with moderate code length and system load factor less than 1.5, the SS-JCAD scheme achieves almost the same bit error rate as the ideal case aided with perfect CSI, implying the SS-JCAD scheme as a near-optimal solution to the massive random access scenario.

Existing search heuristics used to find input values that result in significant floating-point (FP) errors or small ranges that cover them are accompanied by severe constraints, complicating their implementation and restricting their general applicability. This paper introduces an error analysis tool called Eiffel to infer error-inducing input ranges instead of searching them. Given an FP expression with its domain$\mathcal{D}$, Eiffel first constructs an error data set by sampling values across a smaller domain$\mathcal{R}$and assembles these data into clusters. If more than two clusters are formed, Eiffel derives polynomial curves that best fit the bound coordinates of the error-inducing ranges in$\mathcal{R}$, extrapolating them to infer all target ranges of$\mathcal{D}$and reporting the maximal error. Otherwise, Eiffel simply returns the largest error across$\mathcal{R}$. Experimental results show that Eiffel exhibits a broader applicability than Atomu and$\mathbf{S}^{3}$FP by successfully detecting the errors of all 70 considered benchmarks while the two baselines only report errors for part of them. By taking as input the inferred ranges of Eiffel, Herbie obtains an average accuracy improvement of 3.35 bits and up to 53.3 bits.

Abstract With continuous advancements in ancient‐book digitization and preservation research, the problems with the stitching of ancient‐book images have become increasingly prominent, as traditional feature‐mapping‐based methods cannot satisfactorily stitch non‐overlapping images. To realize the accurate stitching of the left and right pages of ancient‐book images, this paper proposes a method for ancient‐book image stitching to meet the requirements of their digitization in back‐wrapped binding and other binding forms. First, a dataset of the black text frames from ancient‐book images was established and then used to train a VGG16‐UNet network for the extraction of black text frames. Then, the Douglas–Peucker algorithm was used to fit the black text frames and filter outliers. Finally, a sliding matching algorithm based on the position information of black text frames was proposed for the rectification of misalignments. The results showed that the method achieved a satisfying stitching effect and had good robustness.

The maximum achievable rate is derived for resistive random-access memory (ReRAM) channel with sneak-path interference. Based on the mutual information spectrum analysis, the maximum achievable rate of ReRAM channel with independent and identically distributed (i.i.d.) binary inputs is derived as an explicit function of channel parameters such as the distribution of cell selector failures and channel noise level. Due to the randomness of cell selector failures, the ReRAM channel demonstrates multi-status characteristic. For each status, it is shown that as the array size is large, the fraction of cells affected by sneak paths approaches a constant value. Therefore, the mutual information spectrum of the ReRAM channel is formulated as a mixture of multiple stationary channels. Maximum achievable rates of the ReRAM channel with different settings, such as single- and across-array codings, with and without data shaping, and optimal and treating-interference-as-noise (TIN) decodings, are compared. These results provide valuable insights on the code design for ReRAM.

With the complication of future communication scenarios, most conventional signal processing technologies of multi-user multiple-input multiple-output (MU-MIMO) become unreliable, which are designed based on ideal assumptions, such as Gaussian signaling and independent identically distributed (IID) channel matrices. As a result, this paper considers a generalized MU-MIMO (GMU-MIMO) system with more general assumptions, i.e., arbitrarily fixed input distributions, and general unitarily-invariant channel matrices. However, there is still no accurate capacity analysis and capacity optimal transceiver with practical complexity for GMU-MIMO under the constraint of coding. To address these issues, inspired by the replica method, the constrained sum capacity of coded GMU-MIMO with fixed input distribution is calculated by using the celebrated mutual information and minimum mean-square error (MMSE) lemma and the MMSE optimality of orthogonal/vector approximate message passing (OAMP/VAMP). Then, a capacity optimal multi-user OAMP/VAMP receiver is proposed, whose achievable rate is proved to be equal to the constrained sum capacity. Moreover, a design principle of multi-user codes is presented for the multi-user OAMP/VAMP, based on which a kind of practical multi-user low-density parity-check (MU-LDPC) code is designed. Numerical results show that finite-length performances of the proposed MU-LDPC codes with multi-user OAMP/VAMP are about 2 dB away from the constrained sum capacity and outperform those of the existing state-of-art methods.

Conventional multi-user multiple-input multiple-output (MU-MIMO) mainly focused on Gaussian signaling, independent and identically distributed (IID) channels, and a limited number of users. It will be laborious to cope with the heterogeneous requirements in next-generation wireless communications, such as various transmission data, complicated communication scenarios, and unprecedented massive user access. Therefore, this paper studies a generalized MU-MIMO (GMU-MIMO) system with more generalized and practical constraints, i.e., practical channel coding, non-Gaussian signaling, right-unitarily-invariant channels (covering Rayleigh fading channel matrices, certain ill-conditioned and correlated channel matrices, etc.), and massive users and antennas. These generalized assumptions bring new challenges in theory and practice. For example, there is no accurate constrained capacity region analysis for GMU-MIMO. In addition, it is unclear how to achieve constrained-capacity-optimal performance with practical complexity. To address these challenges, a unified framework is proposed to derive the constrained capacity region of GMU-MIMO and design a constrained-capacity-optimal transceiver, which jointly considers encoding, modulation, detection, and decoding. Group asymmetry is developed to group users according to their rates, which makes a tradeoff between user rate allocation and implementation complexity. Specifically, the constrained capacity region of group-asymmetric GMU-MIMO is characterized by using the minimum mean-square error (MMSE) optimality of orthogonal/vector approximate message passing (OAMP/VAMP) and the relationship between mutual information and MMSE. Furthermore, a theoretically optimal multi-user OAMP/VAMP receiver and practical multi-user low-density parity-check (MU-LDPC) codes are proposed to achieve the constrained capacity region of group-asymmetric GMU-MIMO. Numerical results demonstrate that the proposed MU-LDPC coded GMU-MIMO systems achieve asymptotic performance within 0.2 dB from the theoretical sum capacity. Moreover, their finite-length performances are about 1~2 dB away from the associated sum capacity of GMU-MIMO.

Resistive random-access memory is one of the most promising candidates for the next generation of non-volatile memory technology. However, its crossbar array structure causes severe “sneak-path” interference, which also leads to strong inter-cell correlation. Recent works have mainly focused on sub-optimal data detection schemes by ignoring inter-cell correlation and assuming sneak-path interference is independent between different array cells. In this paper, we propose a near-optimal data detection scheme that can approach the performance bound of the optimal detection scheme. Our detection scheme leverages a joint data and sneak-path interference recovery and can use all inter-cell correlations. The proposed scheme is suitable for data detection of large memory arrays with only linear operation complexity.

Despite the great promises that the resistive random access memory (ReRAM) has shown as the next generation of non-volatile memory technology, its crossbar array structure leads to a severe sneak path interference to the signal read back from the memory cell. In this paper, we first propose a novel belief propagation (BP) based detector for the sneak path interference in ReRAM. Based on the conditions for a sneak path to occur and the dependence of the states of the memory cells that are involved in the sneak path, a Tanner graph for the ReRAM channel is constructed, inside which specific messages are updated iteratively to get a better estimation of the sneak path affected cells. We further combine the graph of the designed BP detector with that of the BP decoder of the polar codes to form a joint detector and decoder. Tailored for the joint detector and decoder over the ReRAM channel, effective polar codes are constructed using the genetic algorithm. Simulation results show that the BP detector can effectively detect the cells affected by the sneak path, and the proposed polar codes and the joint detector and decoder can significantly improve the error rate performance of ReRAM.

The sneak path (SP) interference problem in resistive random access memory (ReRAM) severely affects the data storage reliability. Recent works showed that the occurrence of the SP is highly related to the selector failures (SFs) in the resistive memory arrays. In this work, we propose a novel scheme to detect the location of the failed selector, based on the signal read back from the memory array. The detected SF location information can be used to assist the data detection to mitigate the SP interference or to construct SP-free constrained codes.

Resistive random-access memory (ReRAM) is a promising candidate for the next generation non-volatile memory technology due to its simple read/write operations and high storage density. However, its crossbar array structure causes a severe interference effect known as the “sneak path.” In this paper, we propose channel coding techniques that can mitigate both the sneak-path interference and the channel noise. The main challenge is that the sneak-path interference is data-dependent, and also correlated within a memory array, and hence the conventional error correction coding scheme will be inadequate. In this work, we propose an across-array coding strategy that assigns a codeword to multiple independent memory arrays, and exploit a real-time channel estimation scheme to estimate the instantaneous status of the ReRAM channel. Since the coded bits from different arrays experience independent channels, a “diversity” gain can be obtained during decoding, and when the codeword is adequately distributed over different memory arrays, the code actually performs as that over an uncorrelated channel. By performing decoding based on the scheme of treating-interference-as-noise (TIN), the ReRAM channel over different memory arrays is equivalent to a block varying channel we defined, for which we propose both the capacity bounds and a coding scheme. The proposed coding scheme consists of a serial concatenation of an optimized error correction code with a data shaper, which enables the ReRAM system to achieve a near capacity limit storage efficiency.

In this paper, we propose channel coding techniques that can mitigate both the sneak-path interference and the channel noise for resistive random-access memory (ReRAM) channels. The main challenge is that the sneak-path interference is data-dependent, and also correlated within a memory array, and hence the conventional error correction coding scheme will be inadequate. We propose an across-array coding scheme, which assigns a code-word to multiple independent memory arrays. Since the coded bits from different arrays experience independent channels, a “diversity” gain can be obtained during decoding, and when the code-word is adequately distributed over different memory arrays, the code actually performs as that over an uncorrelated channel. We also present a real-time channel estimation scheme together with an elementary signal estimator (ESE) to obtain the instant channel status as well as the soft information of the channel coded bits for decoding. By further combining with a data shaping technique to produce an optimized channel input distribution, significant error performance gain is obtained.

The performance of the quasi-cyclic low-density parity-check (QC-LDPC) coded partial-access interleave division multiple access (IDMA) systems is evaluated with the SNR (signal-to-noise ratio) evolution algorithm. The partial access IDMA system is the IDMA system in which the 0s, i.e., non-energy transmission, are inserted into the chip sequence. The SNR evolution algorithm is developed and employed to evaluate the systems. Numerical and simulation results show that the partial access has better BER (bit error rate) performance than that of the conventional full access in a range of low Eb/N0, and the proposed IDMA system with the 3GPP NR QC-LDPC codes has a good error-floor performance.

A very low-complexity on-off division multiple access (ODMA) scheme is proposed for K-user non-orthogonal multiple access (NOMA) systems. At the transmission side, each user employs the same length-m channel code whose coded bits, after modulation, are sent in a random time-hopping manner. Specifically, m coded bits are randomly scheduled and sent using n time slots with n≫m, i.e., only m slots are used for signal transmission and the other n- m slots are idle. The slot selection, referred to as an on-off pattern, is unique to each user, and it is the only means of user separation. Consequently, at each time slot only a very few users (i.e., 2 or 3) may simultaneously access the channel, leading to a super-sparse access system. Due to the sparse access property, a very low-complexity iterative multi-user decoding method can be implemented on an almost tree-like factor graph. Compared with existing iteratively decodable code division multiple access (CDMA) schemes, such as sparse-CDMA and interleave division multiple access (IDMA), ODMA does not rely on repetition (spreading) or user interleaving. In fact, we show that in using extrinsic information transfer (EXIT) analysis and simulation, idling is more effective than repetition in terms of enhancing the multi-user iterative decoding performance. By replacing repetition with idling, a remarkable multi-user decoding performance gain is achieved and, at the same time, the decoding complexity is significantly reduced.

A partial-access scheme is proposed for coded-IDMA (interleave-division multiple-access) systems. Similar to conventional IDMA systems, the transmitter of each user consists of a concatenation of channel encode, spreading, and user-specific interleaving. The only difference is that after spreading, some 0s are inserted into the chip sequence. Among the transmitted symbols of -1, 0, 1, symbol 0, however, implies a non-energy transmission or non-access to the channel. In the IDMA receiver, we use a customary joint iterative decoding based on a single factor graph. In our proposed IDMA systems, partial access not only greatly decreases the complexity of the joint decoding; it also decreases the number of short loops in the factor graph. Our extrinsic information transfer (EXIT) analysis and bit-error-rate (BER) simulations show that the proposed partial-access scheme outperforms conventional full-access-IDMA systems in a low SNR range.

Multilevel flash memories enable multiple bits to be stored in a single memory cell and hence a significant increase of the storage capacity. The multiple bits that are used for labeling the threshold voltage level of a memory cell belong to different pages. A multilevel flash memory channel resembles a multi-user channel with asymmetric noise, while the data of different pages, which are encoded independently, resembles data of multiple users. In this paper, performance analyses are proposed for this channel by using the union bound technique. In particular, we investigated two different binary labeling schemes of a cell level, the Gray labeling and non-Gray labeling with three maximum-likelihood (ML)-based decoding schemes, which are the joint multi-page ML decoding, page-separate ML decoding, and default setting ML decoding. Our analysis reveals an asymptotic diminishing rate of decoding errors as the channel noise approaches zero, based on which the code design criteria are proposed. It is shown theoretically that the Gray mapping has no joint (i.e., multi-page) decoding gain. The corresponding diminishing decoding error rate is dominated by the weakest code in each page and hence a separate decoding scheme is adequate. On the other hand, the non-Gray mapping has a joint decoding gain which means the weak code can exploit the decoding of the strong code and the diminishing rate of its decoding errors is not subject to the cask effect. Therefore, for Gray mapping, a symmetric coding scheme using equal-strength code for each page achieves better error performance, while for non-Gray mapping with joint decoding, the symmetric coding is not necessary. Moreover, by using the asymmetric coding scheme through assigning different code rates to different pages, the non-Gray mapping can achieve higher overall sum rate than Gray mapping with a similar decoding error rate performance.

For the problem of multiple users simultaneously communicating with a single receiver, a sparse multiple access scheme is proposed. Each user employs a low-density parity-check (LDPC) code. To mitigate multi-user interference, the codeword of each user is randomly punctured and the punctured bits are replaced by idle slots. That is, only a small random set of users are active at each time. The restriction of number of concurrent users significantly reduces the multi-user decoding complexity. Moreover, this puncture facilitates an efficient message-passing decoding over a sparse graph. With a joint optimization of the degree distribution of the LDPC code and the column weight distribution of the puncture matrix, capacity-approaching performance is achieved.

MIMO-NOMA combines multiple-input multiple-output (MIMO) and non-orthogonal multiple access (NOMA) techniques to address heterogeneous challenges, such as massive connectivity, low latency, and high reliability in the 5G cellular communication system and beyond. In this paper, a coded MIMO-NOMA system with capacity-approaching performance and low implementation complexity is proposed. Specifically, the proposed MIMO receiver consists of a linear minimum mean-square error (LMMSE) multi-user detector and a bank of single-user message-passing decoders, which decompose the overall NOMA signal recovery into distributed low-complexity computations with iterative processing. An asymptotic extrinsic information transfer analysis is employed to model the overall performance, and a novel class of multi-user irregular repeat-accumulate channel codes that match with the LMMSE multi-user detector in the iterative decoding process are constructed for the system. As a result, the proposed coded MIMO-NOMA system achieves asymptotic performance within 0.2 dB from the theoretical capacity. Simulation results validate the reliability and robustness of the proposed system in practical settings that include different system loads, iteration numbers, code lengths, fast/block fading, and imperfect channel estimation.

In cloud radio access network (C-RAN), remote radio heads (RRHs) and users are uniformly distributed in a large area such that the channel matrix can be considered as sparse. Based on this phenomenon, RRHs only need to detect the relatively strong signals from nearby users and ignore the weak signals from far users, which is helpful to develop low-complexity detection algorithms without causing much performance loss. However, before detection, RRHs require to obtain the realtime user activity information by the dynamic grant procedure, which causes the enormous latency. To address this issue, in this paper, we consider a grant-free C-RAN system and propose a low- complexity Bernoulli-Gaussian message passing (BGMP) algorithm based on the sparsified channel, which jointly detects the user activity and signal. Since active users are assumed to transmit Gaussian signals at any time, the user activity can be regarded as a Bernoulli variable and the signals from all users obey a Bernoulli-Gaussian distribution. In the BGMP, the detection functions for signals are designed with respect to the Bernoulli-Gaussian variable. Numerical results demonstrate the robustness and effectivity of the BGMP. That is, for different sparsified channels, the BGMP can approach the mean-square error (MSE) of the genie-aided sparse minimum mean-square error (GA- SMMSE) which exactly knows the user activity information. Meanwhile, the fast convergence and strong recovery capability for user activity of the BGMP are also verified.

A union bound analysis is given for codes over binary asymmetric channels. By considering a random mapping that modulates each coded bit equiprobably to the signal constellation point, an average union bound is derived explicitly as a function of the code's weight spectrum. The bound can be used for estimating the error floor performance of maximum-likelihood decoding or near optimal decodings.

A union bound and its asymptotic analysis are presented for multilevel flash memory channels. The bound reveals an asymptotic decoding error behaviour under the maximum-likelihood decoding, based on which code design criteria are proposed.

Combination of multiple‐input‐multiple‐output and non‐orthogonal multiple access (MIMO–NOMA) is a promising multiple‐access technology, which can greatly improve spectral efficiency and reduce latency. Among major topics of MIMO–NOMA, an interesting one is how to construct suitable multiuser codes for MIMO–NOMA. In previous works, multiuser codes require to be redesigned when the number of users, transmit antennas, or receive antennas change. Therefore, the previous design methods are too complicated to be applied to MIMO–NOMA. To solve this problem, in this study, the authors first propose a simple multiuser detector (MUD) that detects the signal for each user by regarding the superimposed signal from the other users as interference. Then, based on extrinsic information transfer analysis for the MUD, they propose three criteria to simplify the code design, which copes with the changes of user number and antenna configuration. Moreover, when user number is large, each user requires a low‐rate code to overcome the severe multiuser interference. On the basis of the proposed criteria, they design a low‐rate repetition‐aided irregular repeat‐accumulate (Rep‐IRA) code for MIMO–NOMA with different numbers of users and antennas, which can achieve low complexity with the aid of repetition.

A K ‐user multi‐rate code is proposed for a Gaussian multiple access channel with binary inputs, equal‐power, and symbol synchronisation. In this multi‐rate transmission, K users are equally divided into M groups. For each user in the m th group, a rate‐ regular repeat‐accumulate code serially concatenated with a length‐ spreading is employed. The transmitted rate of each user in the m th group is . At the receiver, iterative joint decoding (IJD) and hybrid interference cancellation (HIC) schemes are considered. For each decoding scheme, a bivariate fixed‐point analysis is applied to explicitly represent as a function of mutual information outputs. On the basis of these basic explicit representations, a united unreliable region is given, where users in at least one group are undecodable. The complementary set of the united unreliable region gives an optimal rate profile that achieves the maximum sum rate. Numerical results show that, for the IJD scheme with M increments, the maximum sum rate increases, approaches the Shannon limit, and exceeds that in conventional equal rate transmission. The maximum sum rate of the HIC scheme, which provides much lower decoding complexity than the IJD scheme, is superior to the conventional successive interference cancellation scheme.

Three potential interference cancellation schemes are compared for the application to a non-orthogonal multiple access communication system. One is the conventional hard successive interference cancellation (SIC) scheme based on independent single-user decodings. The other two, proposed in this paper, are a soft-in soft-out parallel interference cancellation (SISO-PIC) and a hybrid interference cancellation (HIC). The SISO-PIC is an improved joint iterative multi-user detection scheme, which has lower complexity than the prevalent multi-user detection. The HIC combines the advantages of the above two schemes to permit users to be successively processed by a SISO-PIC window according to their receive power levels. A comprehensive comparison is given for these three schemes in aspects of error propagation, detection delay, and complexity when a practical channel code, repeat-accumulate code, is employed. Numerical results show that HIC is a trade-off scheme of the three aspects.

In some image classification tasks, similarities among different categories are different and the samples are usually misclassified as highly similar categories. To distinguish highly similar categories, more specific features are required so that the classifier can improve the classification performance. In this paper, we propose a novel two-level hierarchical feature learning framework based on the deep convolutional neural network (CNN), which is simple and effective. First, the deep feature extractors of different levels are trained using the transfer learning method that fine-tunes the pre-trained deep CNN model toward the new target dataset. Second, the general feature extracted from all the categories and the specific feature extracted from highly similar categories are fused into a feature vector. Then the final feature representation is fed into a linear classifier. Finally, experiments using the Caltech-256, Oxford Flower-102, and Tasmania Coral Point Count (CPC) datasets demonstrate that the expression ability of the deep features resulting from two-level hierarchical feature learning is powerful. Our proposed method effectively increases the classification accuracy in comparison with flat multiple classification methods.

Consider an interleave-division multi-user coding model for a Gaussian multiple-access channel (MAC). Each user's message is encoded by a separate channel encoder followed by a user-interleaver, which is employed for user separation. All the users' codewords are jointly regarded as one multi-user codeword, and the distance between two multi-user codewords is defined as the Euclidean distance after the inter-user superposition caused by the MAC. A distance enumerator analysis estimates the multi-user maximum likelihood decoding performance. By introducing user-scramblers, a multi-user code ensemble is defined, and its distance enumerator is calculated as concatenated multiple single-user codes and a superposition code. Both finite-length and asymptotic analyses are given and it is shown that interleave-division multi-user codes have much better distance properties than conventional random direct-sequence code-division multiple-access. In fact, the codes achieve almost the same asymptotic uniquely decodable performance as a multi-user random code.

A very simple coding scheme, called multi-user repetition-aided irregular repeat-accumulate (IRA) code, is proposed to approach the multiple-access channel (MAC) capacity. The main idea is that not only parity checks, which are generated by an IRA encoder, but also repetitions are used in each user's codeword to reduce the coding and decoding complexities. Repetition is a simple way to construct a low-rate code and is shown to be beneficial for multi-user decoding iteration. It is shown that there is a maximum allowable fraction of repetitions in codewords, below which channel capacity can always be approached by optimizing the degree distribution of the IRA encoder. As users increase, the maximum allowable fraction of repetitions increases, and therefore, very low encoding and decoding complexities are required.

Rate regions of a multi‐way relay channel with direct links (MWRC‐DLs), where K users exchange their messages via a relay terminal and all users can overhear each other directly, are studied in this study. Under the assumption that a restricted encoder is employed at each user, the cut‐set outer bound on the capacity region is derived first. Then, achievable rate regions of the MWRC‐DLs with decode‐and‐forward (DF) and compress‐and‐forward (CF) strategies are characterised. Meanwhile, the explicit expressions of the outer bound and the achievable rate regions for the Gaussian MWRC‐DLs are also derived. It is shown that the rate regions of the DF and CF strategies for the two‐way relay channel and the multiple‐access relay channel can be obtained from those of the MWRC‐DLs. To give more insights on the two strategies of the MWRC‐DLs system, the common rates of a symmetric Gaussian network are analysed. It is shown that the CF strategy achieves common rates within 1/2( K − 1) bits of the capacity when the relay's power is at least ( K − 1) times as large as the user power. Numerical examples are also provided to verify the theoretical analysis.

A K-user nonbinary parallel concatenated code (PCC) is proposed for a Gaussian MAC with symbol synchronization, equal-power, and equal-rate users. In a K-user q-ary PCC over finite field GF(q), each user employs a parallel concatenated code, with a rate-(1/r) q-ary repetition component code and M rate-1 q-ary accumulation component codes. Employing q-ary repetition code is to overcome the multi-user interference and also provide coding gain. The K-user q-ary PCC is not only rate compatible, but also with very low encoding and decoding complexities due to employing such simple component codes. An EXIT chart analysis is given to estimate the decoding threshold of the K-user q-ary PCC. Numerical results show that the decoding threshold of K-user q-ary PCC improves as the field order q increases. The 10-user 64-ary PCC improves the decoding threshold by 1.88 dB over the binary case. The decoding threshold of 15-user 64-ary PCC at sum rate 3/4 is only 0.75 dB away from the Shannon bound. Bit-error-rate simulations are provided to verify the analysis.

A K-user repeat-accumulate interleave-division (RAID) system is considered for a Gaussian multiple access channel (GMAC) with binary inputs, equal-power, and symbol synchronization. In this system, a regular repeat accumulate (RA) code serially concatenated with spreading, where the rate of RA code and spreading length both can be changed, is employed for each user. K users are divided into M groups equally. The mth group contains K/M users with rate Rm(1 ≤ m ≤ M). At the receiver, multiuser message-passing decoding is performed on a single factor graph. A fixed point analysis is developed to obtain all achievable rate profiles for an arbitrary small decoding error rate over the GMAC. It shows that the sum rate of our optimal rate profile is superior to that of conventional multiuser RAID scheme with equal rate.

A distance enumerator analysis is given for multiuser codes over a Gaussian multiple-access channel (MAC). A multi-user code distance is defined as the Euclidean distance between two multi-user codewords after their inter-user super-position caused by MAC. By employing a user-interleaving and a user-scrambling, a multi-user code ensemble is defined, and its distance enumerator is calculated as a concatenated multiple single-user codes and a superposition code.

As a generalization of the binary spreading scheme in conventional direct-sequence code-division multiple-access (DS-CDMA) and interleave-division multiple-access (IDMA), a finite field spreading scheme is proposed for a synchronous multiple-access channel (MAC) with Gaussian noise and equal-power users. For each user, each information symbol over a finite field is spread into a length-L field vector by L-field multiplications. At the receiver, an iterative multi-user decoding algorithm on a factor graph is developed to recover each user's information symbol. To estimate the bit error rate performance of an uncoded finite field spreading system, an extrinsic information transfer analysis of the finite field despreading is given. This analysis shows that in addition to overcoming multi-user interference, the finite field spreading scheme can also provide an additional coding gain to overcome Gaussian noise compared with the conventional spreading scheme. This coding gain increases with the field order. The finite field spreading serially concatenated with a nonbinary low-density parity-check (LDPC) code, with field order 64, approaches the MAC capacity within 0.26 dB at a sum rate of 0.25.

A k-user parallel concatenated code (PCC) is proposed for a Gaussian multiple-access channel with symbol synchronization and equal power users. In this code, each user employs a PCC with M + 1 component codes, where the first component code is a rate 1/q repetition code and the other M component codes are the same rate-1 convolutional code 1/1+D. The K-user PCC achieves a larger maximum sum rate, at the high rate region, than the conventional scheme of an error correction code serially concatenated with a spreading.

As a generalization of the binary coding-spreading scheme, nonbinary coding-spreading scheme is proposed for a synchronous binary-input multiple-access channel (MAC) with Gaussian noise, equal-power, and equal-rate users. In this scheme, each user employs the same nonbinary low-density parity-check code serially concatenated with a nonbinary low-rate mapping, referred to as nonbinary spreading. A user-specific interleaving is employed to make the transmitted data of each user random-like. It is shown that the iterative multi-user decoding threshold of nonbinary coding-spreading scheme is less than 0.5 dB away from the MAC capacity at many sum rates.

A multi-user repeat-accumulate interleave-division (RAID) system is considered for a multiple-access channel (MAC) with binary inputs, equal-power, and symbol synchronization. In the system, a regular repeat-accumulate (RA) code serially concatenated with block spreading is employed for each user. At the receiver, multi-user message-passing decoding is performed on a single factor graph. Over the MAC with additive white Gaussian noise (AWGN), a fixed point analysis is developed to obtain the optimal code rate and the spreading length that give the maximum sum rate for an arbitrary small decoding error rate.