arXiv subjects

Kui Cai

Publications and source records attributed to Kui Cai.

At least 19 recordsLinked to original sources

arXiv2026

Selective Depthwise Separable Convolution for Lightweight Joint Source-Channel Coding in Wireless Image Transmission

Depthwise separable convolutional (DSConv) layers have been successfully applied to deep learning (DL)-based joint source-channel coding (JSCC) schemes to reduce computational complexity. However, a systematic investigation of the layerwise and ratio-wise replacement of standard convolutional (Conv) layers with DSConv layers in JSCC systems for wireless image transmission remains largely unexplored. In this letter, we propose a configurable lightweight JSCC framework that incorporates a selective replacement strategy, enabling flexible Conv-to-DSConv replacement at different replacement ratios and positions. By varying the replacement ratio, we obtain models with different computational complexities and analyze their impact on reconstruction performance. Furthermore, we investigate how replacements at different encoder and decoder depths influence reconstruction quality under a fixed replacement ratio. Our results show that Conv-to-DSConv replacement at the intermediate layers of the encoder and decoder achieves a favorable complexity-performance trade-off, revealing layer-wise redundancy in DL-based JSCC systems. Extensive experiments further demonstrate that the proposed framework achieves substantial parameter reduction with only slight performance degradation, enabling flexible complexity-performance trade-offs for resource-constrained edge devices.

eess.IV↗

arXiv2026

Analog Error Correcting Codes with Constant Redundancy

We consider analog error-correcting codes (analog ECCs) that are designed to correct/detect outlying errors arising in analog implementations of vector-matrix multiplication. The error-correction/detection capability of an analog ECC can be characterized by its height profile, which is expected to be as small as possible. In this paper, we consider analog ECCs whose parity check matrix has columns of unit Euclidean norm. We first present an upper bound on the height profile of such codes as well as a simple decoder for correcting a single error. We then construct a family of single error-correcting analog ECCs with redundancy three for any code length, which has smaller height profile compared to the known $[n,n-2]$ MDS constructions.

cs.IT↗

arXiv2026

U-Net-Based Generative Joint Source-Channel Coding for Wireless Image Transmission

Deep learning (DL)-based joint source-channel coding (JSCC) methods have achieved remarkable success in wireless image transmission. However, these methods either focus on conventional distortion metrics that do not necessarily yield high perceptual quality or incur high computational complexity. In this paper, we propose two DL-based JSCC (DeepJSCC) methods that leverage deep generative architectures for wireless image transmission. Specifically, we propose G-UNet-JSCC, a scheme comprising an encoder and a U-Net-based generator serving as the decoder. Its skip connections enable multi-scale feature fusion to improve both pixel-level fidelity and perceptual quality of reconstructed images by integrating low- and high-level features. To further enhance pixel-level fidelity, the encoder and the U-Net-based decoder are jointly optimized using a weighted sum of structural similarity and mean-squared error (MSE) losses. Building upon G-UNet-JSCC, we further develop a DeepJSCC method called cGAN-JSCC, where the decoder is enhanced through adversarial training. In this scheme, we retain the encoder of G-UNet-JSCC and adversarially train the decoder's generator against a patch-based discriminator. cGAN-JSCC employs a two-stage training procedure. The outer stage trains the encoder and the decoder end-to-end using an MSE loss, while the inner stage adversarially trains the decoder's generator and the discriminator by minimizing a joint loss combining adversarial and distortion losses. Simulation results demonstrate that the proposed methods achieve superior pixel-level fidelity and perceptual quality on both high- and low-resolution images. For low-resolution images, cGAN-JSCC achieves better reconstruction performance and greater robustness to channel variations than G-UNet-JSCC.

eess.IV↗

arXiv2026

On the Sequence Reconstruction Problem for the Single-Deletion Two-Substitution Channel

The Levenshtein sequence reconstruction problem studies the reconstruction of a transmitted sequence from multiple erroneous copies of it. A fundamental question in this field is to determine the minimum number of erroneous copies required to guarantee correct reconstruction of the original sequence. This problem is equivalent to determining the maximum possible intersection size of two error balls associated with the underlying channel. Existing research on the sequence reconstruction problem has largely focused on channels with a single type of error, such as insertions, deletions, or substitutions alone. However, relatively little is known for channels that involve a mixture of error types, for instance, channels allowing both deletions and substitutions. In this work, we study the sequence reconstruction problem for the single-deletion two-substitution channel, which allows one deletion and at most two substitutions applied to the transmitted sequence. Specifically, we prove that if two $q$-ary length-$n$ sequences have the Hamming distance $d\geq 2$, where $q\geq 2$ is any fixed integer, then the intersection size of their error balls under the single-deletion two-substitution channel is upper bounded by $(q^2-1)n^2-(3q^2+5q-5)n+O_q(1)$, where $O_q(1)$ is a constant independent from $n$ but dependent on $q$. Moreover, we show that this upper bound is tight up to an additive constant.

cs.IT↗

arXiv2025

From One-Dimensional Codes to Two-Dimensional Codes: A Universal Framework for the Bounded-Weight Constraint

Recent developments in storage -- especially in the area of resistive random access memory (ReRAM) -- are attempting to scale the storage density by regarding the information data as two-dimensional (2D), instead of one-dimensional (1D). Correspondingly, new types of 2D constraints are introduced into the input information data to improve the system reliability. While 1D constraints have been extensively investigated in the literature, the study for 2D constraints is much less profound. Particularly, given a constraint $\mathcal{F}$ and a design of 1D codes whose codewords satisfy $\mathcal{F}$, the problem of constructing efficient 2D codes, such that every row and every column in every codeword satisfy $\mathcal{F}$, has been a challenge. This work provides an efficient solution to the challenging coding problem above for the binary bounded-weight constrained codes that restrict the maximum number of $1$'s (called {\em weight}). Formally, we propose a universal framework to design 2D codes that guarantee the weight of every row and every column of length $n$ to be at most $f(n)$ for any given function $f(n)$. We show that if there exists a design of capacity-approaching 1D codes, then our method also provides capacity-approaching 2D codes for all $f=\omega(\log n)$.

cs.IT↗

arXiv2025

Sequence Reconstruction for Sticky Insertion/Deletion Channels

The sequence reconstruction problem for insertion/deletion channels has attracted significant attention owing to their applications recently in some emerging data storage systems, such as racetrack memories, DNA-based data storage. Our goal is to investigate the reconstruction problem for sticky-insdel channels where both sticky-insertions and sticky-deletions occur. If there are only sticky-insertion errors, the reconstruction problem for sticky-insertion channel is a special case of the reconstruction problem for tandem-duplication channel which has been well-studied. In this work, we consider the $(t, s)$-sticky-insdel channel where there are at most $t$ sticky-insertion errors and $s$ sticky-deletion errors when we transmit a message through the channel. For the reconstruction problem, we are interested in the minimum number of distinct outputs from these channels that are needed to uniquely recover the transmitted vector. We first provide a recursive formula to determine the minimum number of distinct outputs required. Next, we provide an efficient algorithm to reconstruct the transmitted vector from erroneous sequences.

cs.IT↗

arXiv2025

Constrained Coding for Composite DNA: Channel Capacity and Efficient Constructions

Composite DNA is a recent novel method to increase the information capacity of DNA-based data storage above the theoretical limit of 2 bits/symbol. In this method, every composite symbol does not store a single DNA nucleotide but a mixture of the four nucleotides in a predetermined ratio. By using different mixtures and ratios, the alphabet can be extended to have much more than four symbols in the naive approach. While this method enables higher data content per synthesis cycle, potentially reducing the DNA synthesis cost, it also imposes significant challenges for accurate DNA sequencing since the base-level errors can easily change the mixture of bases and their ratio, resulting in changes to the composite symbols. With this motivation, we propose efficient constrained coding techniques to enforce the biological constraints, including the runlength-limited constraint and the GC-content constraint, into every DNA synthesized oligo, regardless of the mixture of bases in each composite letter and their corresponding ratio. Our goals include computing the capacity of the constrained channel, constructing efficient encoders/decoders, and providing the best options for the composite letters to obtain capacity-approaching codes. For certain codes' parameters, our methods incur only one redundant symbol.

cs.IT↗

arXiv2025

Sequence Reconstruction for the Single-Deletion Single-Substitution Channel

The central problem in sequence reconstruction is to find the minimum number of distinct channel outputs required to uniquely reconstruct the transmitted sequence. According to Levenshtein's work in 2001, this number is determined by the size of the maximum intersection between the error balls of any two distinct input sequences of the channel. In this work, we study the sequence reconstruction problem for single-deletion single-substitution channel, assuming that the transmitted sequence belongs to a $q$-ary code with minimum Hamming distance at least $2$, where $q\geq 2$ is any fixed integer. Specifically, we prove that for any two $q$-ary sequences of length $n$ and with Hamming distance $d\geq 2$, the size of the intersection of their error balls is upper bounded by $2qn-3q-2-\delta_{q,2}$, where $\delta_{i,j}$ is the Kronecker delta. We also prove the tightness of this bound by constructing two sequences the intersection size of whose error balls achieves this bound.

cs.IT↗

arXiv2024

Performance Analysis and Code Design for Resistive Random-Access Memory Using Channel Decomposition Approach

A novel framework for performance analysis and code design is proposed to address the sneak path (SP) problem in resistive random-access memory (ReRAM) arrays. The main idea is to decompose the ReRAM channel, which is both non-ergodic and data-dependent, into multiple stationary memoryless channels. A finite-length performance bound is derived by analyzing the capacity and dispersion of these stationary memoryless channels. Furthermore, leveraging this channel decomposition, a practical sparse-graph code design is proposed using density evolution. The obtained channel codes are not only asymptotic capacity approaching but also close to the derived finite-length performance bound.

cs.IT↗

arXiv2024

Constrained Coding and Deep Learning Aided Threshold Detection for Resistive Memories

Resistive random access memory (ReRAM) is a promising emerging non-volatile memory (NVM) technology that shows high potential for both data storage and computing. However, its crossbar array architecture leads to the sneak path problem, which may severely degrade the reliability of data stored in the ReRAM cell. Due to the complication of memory physics and unique features of the sneak path induced interference (SPI), it is difficult to derive an accurate channel model for it. The deep learning (DL)-based detection scheme \cite{zhong2020sneakdl} can better mitigate the SPI, at the cost of additional power consumption and read latency. In this letter, we first propose a novel CC scheme which can not only reduce the SPI in the memory array, but also effectively differentiate the memory arrays into two categories of sneak-path-free and sneak-path-affected arrays. For the sneak-path-free arrays, we can use a simple middle-point threshold detector to detect the low and high resistance cells of ReRAM. For the sneak-path-affected arrays, a DL detector is first trained off-line (prior to the data detection of ReRAM). To avoid the additional power consumption and latency introduced by the DL detector, we further propose a DL-based threshold detector, whose detection threshold can be derived based on the outputs of the DL detector. It is then utilized for the online data detection of all the identified sneak-path-affected arrays. Simulation results demonstrate that the above CC and DL aided threshold detection scheme can effectively mitigate the SPI of the ReRAM array and achieve better error rate performance than the prior art detection schemes, without the prior knowledge of the channel.

cs.IT↗

arXiv2024

Deep Learning-Based Decoding of Linear Block Codes for Spin-Torque Transfer Magnetic Random Access Memory (STT-MRAM)

Thanks to its superior features of fast read/write speed and low power consumption, spin-torque transfer magnetic random access memory (STT-MRAM) has become a promising non-volatile memory (NVM) technology that is suitable for many applications. However, the reliability of STT-MRAM is seriously affected by the variation of the memory fabrication process and the working temperature, and the later will lead to an unknown offset of the channel. Hence, there is a pressing need to develop more effective error correction coding techniques to tackle these imperfections and improve the reliability of STT-MRAM. In this work, we propose, for the first time, the application of deep-learning (DL) based algorithms and techniques to improve the decoding performance of linear block codes with short codeword lengths for STT-MRAM. We formulate the belief propagation (BP) decoding of linear block code as a neural network (NN), and propose a novel neural normalized-offset reliability-based min-sum (NNORB-MS) decoding algorithm. We successfully apply our proposed decoding algorithm to the STT-MRAM channel through channel symmetrization to overcome the channel asymmetry. We also propose an NN-based soft information generation method (SIGM) to take into account the unknown offset of the channel. Simulation results demonstrate that our proposed NNORB-MS decoding algorithm can achieve significant performance gain over both the hard-decision decoding (HDD) and the regular reliability-based min-sum (RB-MS) decoding algorithm, for cases without and with the unknown channel offset. Moreover, the decoder structure and time complexity of the NNORB-MS algorithm remain similar to those of the regular RB-MS algorithm.

cs.IT↗

arXiv2024

Deep Transfer Learning-based Detection for Flash Memory Channels

The NAND flash memory channel is corrupted by different types of noises, such as the data retention noise and the wear-out noise, which lead to unknown channel offset and make the flash memory channel non-stationary. In the literature, machine learning-based methods have been proposed for data detection for flash memory channels. However, these methods require a large number of training samples and labels to achieve a satisfactory performance, which is costly. Furthermore, with a large unknown channel offset, it may be impossible to obtain enough correct labels. In this paper, we reformulate the data detection for the flash memory channel as a transfer learning (TL) problem. We then propose a model-based deep TL (DTL) algorithm for flash memory channel detection. It can effectively reduce the training data size from $10^6$ samples to less than 104 samples. Moreover, we propose an unsupervised domain adaptation (UDA)-based DTL algorithm using moment alignment, which can detect data without any labels. Hence, it is suitable for scenarios where the decoding of error-correcting code fails and no labels can be obtained. Finally, a UDA-based threshold detector is proposed to eliminate the need for a neural network. Both the channel raw error rate analysis and simulation results demonstrate that the proposed DTL-based detection schemes can achieve near-optimal bit error rate (BER) performance with much less training data and/or without using any labels.

cs.IT↗

arXiv2024

Sneak Path Interference-Aware Adaptive Detection and Decoding for Resistive Memory Arrays

Resistive random-access memory (ReRAM) is an emerging non-volatile memory technology for high-density and high-speed data storage. However, the sneak path interference (SPI) occurred in the ReRAM crossbar array seriously affects its data recovery performance. In this letter, we first propose a quantized channel model of ReRAM, based on which we design both the one-bit and multi-bit channel quantizers by maximizing the mutual information of the channel. A key channel parameter that affects the quantizer design is the sneak path occurrence probability (SPOP) of the memory cell. We first use the average SPOP calculated statistically to design the quantizer, which leads to the same channel detector for different memory arrays. We then adopt the SPOP estimated separately for each memory array for the quantizer design, which is generated by an effective channel estimator and through an iterative detection and decoding scheme for the ReRAM channel. This results in an array-level SPI-aware adaptive detection and decoding approach. Moreover, since there is a strong correlation of the SPI that affects memory cells in the same rows/columns than that affecting cells in different rows/columns, we further derive a column-level scheme which outperforms the array-level scheme. We also propose a channel decomposition method that enables effective ways for theoretically analyzing the ReRAM channel. Simulation results show that the proposed SPI-aware adaptive detection and decoding schemes can approach the ideal performance with three quantization bits, with only one decoding iteration.

cs.IT↗

arXiv2024

Maximum Achievable Rate of Resistive Random-Access Memory Channels by Mutual Information Spectrum Analysis

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.

cs.IT↗

arXiv2024

High Information Density and Low Coverage Data Storage in DNA with Efficient Channel Coding Schemes

DNA-based data storage has been attracting significant attention due to its extremely high data storage density, low power consumption, and long duration compared to conventional data storage media. Despite the recent advancements in DNA data storage technology, significant challenges remain. In particular, various types of errors can occur during the processes of DNA synthesis, storage, and sequencing, including substitution errors, insertion errors, and deletion errors. Furthermore, the entire oligo may be lost. In this work, we report a DNA-based data storage architecture that incorporates efficient channel coding schemes, including different types of error-correcting codes (ECCs) and constrained codes, for both the inner coding and outer coding for the DNA data storage channel. We also carried out large scale experiments to validate our proposed DNA-based data storage architecture. Specifically, 1.61 and 1.69 MB data were encoded into 30,000 oligos each, with information densities of 1.731 and 1.815, respectively. It has been found that the stored information can be fully recovered without any error at average coverages of 4.5 and 6.0, respectively. This experiment achieved the highest net information density and lowest coverage among existing DNA-based data storage experiments (with standard DNA), with data recovery rates and coverage approaching theoretical optima.

cs.IT↗

arXiv2024

Quantization Design for Resistive Memories With Multiple Reads

Due to the crossbar array architecture, the sneak-path problem severely degrades the data integrity in the resistive random access memory (ReRAM). In this letter, we investigate the channel quantizer design for ReRAM arrays with multiple reads, which is a typical technique to improve the data recovery performance of data storage systems. Starting with a quantized channel model of ReRAM with multiple reads, we first derive a general approach for designing the channel quantizer, for both single-bit and multiple-bit quantization. We then focus on the single-bit quantization, which is highly suitable for practical applications of ReRAM. In particular, we propose a semi-analytical approach to design the multiple-read single-bit quantizer with less complexity. We also derive the theoretical bit-error probability of the optimal single-bit detector/quantization as the benchmark. Results indicate that the multiple-read operation is effective in improving the error rate performance of ReRAM. Moreover, our proposed multiple-read detector outperforms the prior art detector and achieves the performance of the optimal detector.

cs.IT↗

arXiv2024

Union Bound Analysis for Spin-Torque Transfer Magnetic Random Access Memory (STT-MRAM) With Channel Quantization

As an emerging non-volatile memory (NVM) technology, spin-torque transfer magnetic random access memory (STT-MRAM) has received great attention in recent years since it combines the features of low switching energy, fast write/read speed, and high scalability. However, process variation and thermal fluctuation severely affect the data integrity of STT-MRAM, resulting in both write errors and read errors. Therefore, effective error correction codes (ECCs) are necessary for correcting memory cell errors. Meanwhile, the design of channel quantizer plays a critical role in supporting error correction coding for STT-MRAM. In this work, we propose a union bound analysis which can accurately predict the word error rates (WERs) of ECCs with maximum-likelihood (ML) decoding over the quantized STT-MRAM channel. The derived bound provides a theoretical tool for comparing the performance of ECCs with different quantization schemes at very low error rate levels without resorting to lengthy computer simulations. Moreover, we also propose a new criterion to design the channel quantizer by minimizing the WERs of ECC decoding that are obtained from the union bound analysis. Numerical results show that the proposed union-bound-optimized (UBO) quantizer can achieve better error rate performance than the state-of-art quantizers for STT-MRAM.

cs.IT↗

arXiv2024

Deep-Learning-Based Adaptive Error-Correction Decoding for Spin-Torque Transfer Magnetic Random Access Memory (STT-MRAM)

Spin-torque transfer magnetic random access memory (STT-MRAM) is a promising emerging non-volatile memory (NVM) technology with wide applications. However, the data recovery of STT-MRAM is affected by the diversity of channel raw bit error rate (BER) across different dies caused by process variations, as well as the unknown resistance offset due to temperature change. Therefore, it is critical to develop effective decoding algorithms of error correction codes (ECCs) for STT-MRAM. In this article, we first propose a neural bit-flipping (BF) decoding algorithm, which can share the same trellis representation as the state-of-the-art neural decoding algorithms, such as the neural belief propagation (NBP) and neural offset min-sum (NOMS) algorithm. Hence, a neural network (NN) decoder with a uniform architecture but different NN parameters can realize all these neural decoding algorithms. Based on such a unified NN decoder architecture, we further propose a novel deep-learning (DL)-based adaptive decoding algorithm whose decoding complexity can be adjusted according to the change of the channel conditions of STT-MRAM. Extensive experimental evaluation results demonstrate that the proposed neural decoders can greatly improve the performance over the standard decoders, with similar decoding latency and energy consumption. Moreover, the DL-based adaptive decoder can work well over different channel conditions of STT-MRAM irrespective of the unknown resistance offset, with a 50% reduction of the decoding latency and energy consumption compared to the fixed decoder.

cs.IT↗