迁移学习赋能CNN：面向ReRAM潜路径干扰的高效数据检测方法

戴彬; 吴安妮

doi:10.11999/JEIT260354

迁移学习赋能CNN：面向ReRAM潜路径干扰的高效数据检测方法

doi: 10.11999/JEIT260354 cstr: 32379.14.JEIT260354

戴彬^,,
吴安妮

南京邮电大学南京 210003

基金项目: 国家自然科学基金(62201283)

详细信息

作者简介:
戴彬：男，副教授，研究方向为无线通信与存储系统中的编码与信号处理

吴安妮：女，硕士生，研究方向为存储系统中的信号处理

通讯作者:
戴彬　daibin@njupt.edu.cn

中图分类号: TN492; TP301.1
计量
- 文章访问数: 442
- HTML全文浏览量: 209
- PDF下载量: 16
- 被引次数: 0
出版历程
- 收稿日期: 2026-03-26
- 修回日期: 2026-05-15
- 录用日期: 2026-05-15
- 网络出版日期: 2026-06-02

Transfer Learning Aided CNN for Efficient Data Detection in ReRAM with Sneak-Path Interference

DAI Bin^,,
WU Anni

Nanjing University of Posts and Telecommunications, Nanjing 210003, China

Funds: The National Natural Science Foundation of China (62201283)

摘要

摘要: 阻变存储器中的SPI会引发不可预测的单元间相关性，显著提高信号检测的复杂度。传统检测方法通常依赖已知的信道噪声状态假设，在实际应用中泛化能力有限。为解决这一问题，该文提出了3种基于卷积神经网络(CNN)的数据检测方法，无须依赖先验信道信息即可有效建模并抑制干扰：其一，结合约束编码与多层CNN，通过约束编码判断潜路径干扰状态并进行数据恢复；其二，采用双CNN框架，首先利用轻量CNN进行SPI识别，再通过多层CNN实现精细化检测；其三，引入迁移学习机制，在保持检测精度的同时，将所需训练样本规模降低至传统方法的千分之一。仿真结果表明，所提方法在未知信道状态下均实现误码率性能的显著提升，误比特率相比现有算法至少降低一半，逼近性能极限。此外，迁移学习机制的引入将训练样本需求由$ {10}^{6} $组降至$ 1000 $组，减少3个数量级。
- 阻变存储器 /
- SPI /
- 数据检测 /
- 卷积神经网络 /
- 迁移学习
Abstract: Objective Sneak Path Interference (SPI) in Resistive Random-Access Memory (ReRAM) introduces unpredictable inter-cell correlations and increases the complexity of signal detection. Conventional detection methods usually rely on known channel noise states, which limits their generalization in practical applications. To address this issue, three data detection methods based on Convolutional Neural Networks (CNNs) are proposed to model and suppress interference without prior channel information. The first method combines constrained coding with a multilayer CNN to determine the SPI state and recover data. The second method uses a dual-CNN framework, in which a lightweight CNN first identifies SPI and a multilayer CNN then performs refined detection. The third method incorporates Transfer Learning (TL) to maintain detection accuracy while reducing the required training sample size to one thousandth of that required by conventional methods. Simulation results show that the proposed methods achieve improved Bit Error Rate (BER) performance under unknown channel conditions. Compared with existing algorithms, the BER is reduced by at least half and approaches the theoretical performance limit. In addition, TL reduces the required number of training samples from$ {10}^{6} $ to $ 1\;000 $, corresponding to a reduction of three orders of magnitude. Methods To address different challenges in SPI detection, this paper proposes three methods. First, the Constrained Coding-Aided CNN (CC-CNN) detection framework is developed to handle complex inter-cell correlations caused by SPI. This method first uses constrained coding to detect the presence of interference. It then uses a CNN to learn random correlations under interference and achieve accurate signal recovery. Second, the dual-CNN-based detection method is proposed to avoid the code-rate loss caused by conventional constrained coding. By directly using a CNN to learn and identify SPI patterns from raw data, this method removes the need for redundant coding or extra overhead. It achieves high-precision interference detection while preserving the overall code-rate performance of the system. Third, the TL-based CNN (TL-CNN) detection method is designed to reduce the dependence of high-performance CNNs on large training datasets. By reusing knowledge from pretrained models, this method enables rapid adaptation to ReRAM signal detection tasks. It reduces the required number of training samples while maintaining high detection accuracy and resource efficiency, thereby improving its feasibility in practical scenarios. Results and Discussions Simulation results show that the three proposed methods consistently approach the theoretical lower bound (Fig. 4). They outperform baseline methods, including the Belief Propagation (BP) detector, Deep Neural Network (DNN) detector, and Elementary Signal Estimator (ESE) detector. The two-step network achieves performance comparable to that of the single-step network while avoiding code-rate loss. The TL-CNN detector attains near-optimal BER with only 1,000 target-domain samples, and its performance stabilizes when the sample size exceeds 1,000 (Fig. 6), confirming its data efficiency. The integration of the Selective Kernel (SK) module enables the models to capture SPI-induced spatial correlations effectively (Fig. 5). The TL strategy also supports robust performance under different noise conditions. Conclusions The crossbar array architecture of ReRAM is susceptible to SPI during storage operations, reducing data reliability. To address this issue, this paper proposes three deep learning-based detection methods. Type-I integrates constrained coding with a CNN for efficient and fast interference detection. Type-II uses a two-stage processing strategy. It first classifies interference patterns in the memory array and then detects affected units, thereby maintaining high detection accuracy while minimizing code-rate loss. Type-III introduces a TL framework that uses a pretrained source-domain model to reduce the number of target-domain training samples and lower training overhead. Experimental results show that, under different noise conditions, all three proposed methods achieve performance close to the theoretical lower bound. These methods provide an effective solution for improving the reliability of ReRAM storage systems.
- ReRAM /
- Sneak path interference /
- Data detection /
- Convolutional neural network /
- Transfer learning

HTML全文

图 1 4×4 ReRAM阵列中SPI的示意图

下载: 全尺寸图片幻灯片

图 2 多层选择性卷积核CNN网络的架构图

下载: 全尺寸图片幻灯片

图 3 双卷积神经网络辅助数据检测

下载: 全尺寸图片幻灯片

图 4 不同检测器的误比特率性能对比

下载: 全尺寸图片幻灯片

图 5 CNN网络消融实验的误比特率性能对比

下载: 全尺寸图片幻灯片

图 6 迁移学习目标域中不同样本数量下的误比特率性能对比

下载: 全尺寸图片幻灯片

表 1 测试时间对比

算法 DNN检测 BP检测 BiLSTM检测 Type-I/II检测(无SK) Type-I/II检测(有SK)

时间(s) 15 99 21 10/17 34/43

下载: 导出CSV

参考文献(28)

[1]	STRUKOV D B, SNIDER G S, STEWART D R, et al. The missing memristor found[J]. Nature, 2008, 453(7191): 80–83. doi: 10.1038/nature06932.
[2]	李源堃, 王泽, 张清天, 等. NAS4CIM: 面向忆阻器存算一体芯片的神经网络结构搜索框架[J]. 电子与信息学报, 2025, 47(12): 4948–4958. doi: 10.11999/JEIT250978. LI Yuankun, WANG Ze, ZHANG Qingtian, et al. NAS4CIM: Tailored neural network architecture search for RRAM-based compute-in-memory chips[J]. Journal of Electronics & Information Technology, 2025, 47(12): 4948–4958. doi: 10.11999/JEIT250978.
[3]	李付鹏, 王光义, 刘敬彪, 等. 局部有源忆阻器电路的类生物神经网络存算研究[J]. 电子与信息学报, 2026, 48(2): 866–872. doi: 10.11999/JEIT250631. LI Fupeng, WANG Guangyi, LIU Jingbiao, et al. The storage and calculation of biological-like neural networks for locally active memristor circuits[J]. Journal of Electronics & Information Technology, 2026, 48(2): 866–872. doi: 10.11999/JEIT250631.
[4]	刘嵩, 李子涵, 邱达, 等. 可控多双涡卷忆阻Hopfield神经网络建模及其动力学分析[J]. 电子与信息学报, 2026, 48(1): 417–428. doi: 10.11999/JEIT250972. LIU Song, LI Zihan, QIU Da, et al. Modeling and dynamic analysis of controllable multi-double scroll memristor Hopfield neural network[J]. Journal of Electronics & Information Technology, 2026, 48(1): 417–428. doi: 10.11999/JEIT250972.
[5]	RAJENDRAN G, BASAK D, DEB S, et al. Securing binarized neural networks via PUF-based key management in memristive crossbar arrays[J]. IEEE Embedded Systems Letters, 2025, 17(1): 30–33. doi: 10.1109/LES.2024.3422294.
[6]	蔺海荣, 段晨星, 邓晓衡, 等. 双忆阻类脑混沌神经网络及其在IoMT数据隐私保护中应用[J]. 电子与信息学报, 2025, 47(7): 2194–2210. doi: 10.11999/JEIT241133. LIN Hairong, DUAN Chenxing, DENG Xiaoheng, et al. Dual-memristor brain-like chaotic neural network and its application in IoMT data privacy protection[J]. Journal of Electronics & Information Technology, 2025, 47(7): 2194–2210. doi: 10.11999/JEIT241133.
[7]	ZIDAN M A, FAHMY H A H, HUSSAIN M M, et al. Memristor-based memory: The sneak paths problem and solutions[J]. Microelectronics Journal, 2013, 44(2): 176–183. doi: 10.1016/j.mejo.2012.10.001.
[8]	WANG T J, LI Chengying, SHIH P A, et al. Development of fully ZnO-based 16 × 16 1S1R RRAM crossbar array and performance investigations[J]. IEEE Transactions on Electron Devices, 2025, 72(4): 1702–1708. doi: 10.1109/TED.2025.3539650.
[9]	BAO Shengyu, WANG Zongwei, YANG Yuhang, et al. Advancing toward 4F² 1T1R RRAM with local NAND-gate and isolation scheme[J]. IEEE Transactions on Electron Devices, 2025, 72(5): 2327–2333. doi: 10.1109/TED.2025.3554165.
[10]	BEN-HUR Y and CASSUTO Y. Detection and coding schemes for sneak-path interference in resistive memory arrays[J]. IEEE Transactions on Communications, 2019, 67(6): 3821–3833. doi: 10.1109/TCOMM.2019.2897762.
[11]	CHEN Zehui, SCHOENY C, and DOLECEK L. Pilot assisted adaptive thresholding for sneak-path mitigation in resistive memories with failed selection devices[J]. IEEE Transactions on Communications, 2020, 68(1): 66–81. doi: 10.1109/TCOMM.2019.2948332.
[12]	SONG Guanghui, CAI Kui, ZHONG Xingwei, et al. Performance limit and coding schemes for resistive random-access memory channels[J]. IEEE Transactions on Communications, 2021, 69(4): 2093–2106. doi: 10.1109/TCOMM.2021.3051413.
[13]	LI Panpan, CAI Kui, SONG Guanghui, et al. Sneak Path interference-aware adaptive detection and decoding for resistive memory arrays[J]. IEEE Communications Letters, 2022, 26(9): 2032–2036. doi: 10.1109/LCOMM.2022.3186324.
[14]	SUN Ce, CAI Kui, SONG Guanghui, et al. Belief propagation based joint detection and decoding for resistive random access memories[J]. IEEE Transactions on Communications, 2022, 70(4): 2227–2239. doi: 10.1109/TCOMM.2022.3156074.
[15]	SONG Guanghui, CAI Kui, SUN Ce, et al. Near-optimal detection for both data and sneak-path interference in resistive memories with random cell selector failures[J]. IEEE Transactions on Communications, 2022, 70(2): 836–850. doi: 10.1109/TCOMM.2022.3140459.
[16]	ZHONG Xingwei, CAI Kui, SONG Guanghui, et al. Constrained coding and deep learning aided threshold detection for resistive memories[J]. IEEE Communications Letters, 2022, 26(4): 803–807. doi: 10.1109/LCOMM.2022.3148292.
[17]	MEI Zhen, JU Minghui, CAI Kui, et al. Meta-transfer learning-based few-shot data detection for resistive memory channels[C]. 2025 IEEE Information Theory Workshop (ITW), Sydney, Australia, 2025: 638–643. doi: 10.1109/ITW62417.2025.11240479.
[18]	LI Panpan, SONG Guanghui, CAI Kui, et al. Across-array coding for resistive memories with sneak-path interference and lognormal distributed resistance variations[J]. IEEE Communications Letters, 2021, 25(11): 3458–3462. doi: 10.1109/LCOMM.2021.3111218.
[19]	DAI Bin, CAI Kui, MEI Zhen, et al. Polar code construction for resistive memories with sneak-path interference[J]. IEEE Communications Letters, 2024, 28(8): 1765–1769. doi: 10.1109/LCOMM.2024.3413754.
[20]	PANG Qike, MA Zheng, and TANG Xiaohu. Unreliability normalization weighted bit-flipping algorithms of LDPC decoding for ReRAM systems[J]. Science China Information Sciences, 2024, 67(12): 229304. doi: 10.1007/s11432-024-4228-3.
[21]	KONG Lingjun, QI Yingnan, LIU Haiyang, et al. Sneak path-aware reliability-based iterative majority-logic decoding algorithms for LDPC codes in ReRAM systems[J]. IEEE Communications Letters, 2025, 29(9): 2018–2022. doi: 10.1109/LCOMM.2025.3584075.
[22]	SONG Guanghui, GAO Meiru, LI Ying, et al. Performance analysis and code design for resistive random-access memory using channel decomposition approach[J]. IEEE Transactions on Information Theory, 2026, 72(1): 358–373. doi: 10.1109/TIT.2025.3638148.
[23]	SHI Zhifang, FANG Yi, BU Yingcheng, et al. Convolutional neural network (CNN)-based detection for multi-level-cell NAND flash memory[J]. IEEE Communications Letters, 2021, 25(12): 3883–3887. doi: 10.1109/LCOMM.2021.3112908.
[24]	ZHANG Lei and GAO Xinbo. Transfer adaptation learning: A decade survey[J]. IEEE Transactions on Neural Networks and Learning Systems, 2024, 35(1): 23–44. doi: 10.1109/TNNLS.2022.3183326.
[25]	MEI Zhen, CAI Kui, SHI Long, et al. Deep transfer learning-based detection for flash memory channels[J]. IEEE Transactions on Communications, 2024, 72(6): 3425–3438. doi: 10.1109/TCOMM.2024.3357616.
[26]	SONG Guanghui, WANG Junqi, GAO Meiru, et al. Probability distribution of sneak path rate in resistive random-access memory arrays[C]. 2025 IEEE International Symposium on Information Theory (ISIT), Ann Arbor, USA, 2025: 1–6. doi: 10.1109/ISIT63088.2025.11195609.
[27]	LI Xiang, WANG Wenhai, HU Xiaolin, et al. Selective kernel networks[C]. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, USA, 2019: 510–519. doi: 10.1109/CVPR.2019.00060.
[28]	NGUYEN C T, VAN HUYNH N, CHU N H, et al. Transfer learning for wireless networks: A comprehensive survey[J]. Proceedings of the IEEE, 2022, 110(8): 1073–1115. doi: 10.1109/JPROC.2022.3175942.