• 文献检索
  • 文档翻译
  • 深度研究
  • 学术资讯
  • Suppr Zotero 插件Zotero 插件
  • 邀请有礼
  • 套餐&价格
  • 历史记录
应用&插件
Suppr Zotero 插件Zotero 插件浏览器插件Mac 客户端Windows 客户端微信小程序
定价
高级版会员购买积分包购买API积分包
服务
文献检索文档翻译深度研究API 文档MCP 服务
关于我们
关于 Suppr公司介绍联系我们用户协议隐私条款
关注我们

Suppr 超能文献

核心技术专利：CN118964589B侵权必究
粤ICP备2023148730 号-1Suppr @ 2026

文献检索

告别复杂PubMed语法，用中文像聊天一样搜索，搜遍4000万医学文献。AI智能推荐，让科研检索更轻松。

立即免费搜索

文件翻译

保留排版，准确专业，支持PDF/Word/PPT等文件格式，支持 12+语言互译。

免费翻译文档

深度研究

AI帮你快速写综述，25分钟生成高质量综述，智能提取关键信息，辅助科研写作。

立即免费体验

具有非线性权重更新的忆阻器突触的脉冲神经网络（SNN）

Spiking Neural Network (SNN) With Memristor Synapses Having Non-linear Weight Update.

作者信息

Kim Taeyoon, Hu Suman, Kim Jaewook, Kwak Joon Young, Park Jongkil, Lee Suyoun, Kim Inho, Park Jong-Keuk, Jeong YeonJoo

机构信息

Center for Neuromorphic Engineering, Korea Institutes of Science and Technology, Seoul, South Korea.

出版信息

Front Comput Neurosci. 2021 Mar 11;15:646125. doi: 10.3389/fncom.2021.646125. eCollection 2021.

DOI:10.3389/fncom.2021.646125
PMID:33776676
原文链接:https://pmc.ncbi.nlm.nih.gov/articles/PMC7996210/
Abstract

Among many artificial neural networks, the research on Spike Neural Network (SNN), which mimics the energy-efficient signal system in the brain, is drawing much attention. Memristor is a promising candidate as a synaptic component for hardware implementation of SNN, but several non-ideal device properties are making it challengeable. In this work, we conducted an SNN simulation by adding a device model with a non-linear weight update to test the impact on SNN performance. We found that SNN has a strong tolerance for the device non-linearity and the network can keep the accuracy high if a device meets one of the two conditions: 1. symmetric LTP and LTD curves and 2. positive non-linearity factors for both LTP and LTD. The reason was analyzed in terms of the balance between network parameters as well as the variability of weight. The results are considered to be a piece of useful prior information for the future implementation of emerging device-based neuromorphic hardware.

摘要

在众多人工神经网络中，对模仿大脑中节能信号系统的脉冲神经网络（SNN）的研究备受关注。忆阻器作为用于SNN硬件实现的突触组件是一个有前景的候选者，但一些非理想的器件特性使其具有挑战性。在这项工作中，我们通过添加一个具有非线性权重更新的器件模型进行了SNN模拟，以测试其对SNN性能的影响。我们发现SNN对器件非线性具有很强的耐受性，并且如果器件满足以下两个条件之一，网络就能保持高精度：1. 对称的长时程增强（LTP）和长时程抑制（LTD）曲线；2. LTP和LTD的正非线性因子。从网络参数之间的平衡以及权重的可变性方面分析了原因。这些结果被认为是未来基于新兴器件的神经形态硬件实现的有用先验信息。

https://cdn.ncbi.nlm.nih.gov/pmc/blobs/d605/7996210/d91ade4d2589/fncom-15-646125-g0007.jpg
https://cdn.ncbi.nlm.nih.gov/pmc/blobs/d605/7996210/4efef3a5c84d/fncom-15-646125-g0001.jpg
https://cdn.ncbi.nlm.nih.gov/pmc/blobs/d605/7996210/fc24b0b37b5a/fncom-15-646125-g0002.jpg
https://cdn.ncbi.nlm.nih.gov/pmc/blobs/d605/7996210/a719d7d3b0c5/fncom-15-646125-g0003.jpg
https://cdn.ncbi.nlm.nih.gov/pmc/blobs/d605/7996210/b297395194cc/fncom-15-646125-g0004.jpg
https://cdn.ncbi.nlm.nih.gov/pmc/blobs/d605/7996210/f3161a933a7d/fncom-15-646125-g0005.jpg
https://cdn.ncbi.nlm.nih.gov/pmc/blobs/d605/7996210/d3bc2be93564/fncom-15-646125-g0006.jpg
https://cdn.ncbi.nlm.nih.gov/pmc/blobs/d605/7996210/d91ade4d2589/fncom-15-646125-g0007.jpg
https://cdn.ncbi.nlm.nih.gov/pmc/blobs/d605/7996210/4efef3a5c84d/fncom-15-646125-g0001.jpg
https://cdn.ncbi.nlm.nih.gov/pmc/blobs/d605/7996210/fc24b0b37b5a/fncom-15-646125-g0002.jpg
https://cdn.ncbi.nlm.nih.gov/pmc/blobs/d605/7996210/a719d7d3b0c5/fncom-15-646125-g0003.jpg
https://cdn.ncbi.nlm.nih.gov/pmc/blobs/d605/7996210/b297395194cc/fncom-15-646125-g0004.jpg
https://cdn.ncbi.nlm.nih.gov/pmc/blobs/d605/7996210/f3161a933a7d/fncom-15-646125-g0005.jpg
https://cdn.ncbi.nlm.nih.gov/pmc/blobs/d605/7996210/d3bc2be93564/fncom-15-646125-g0006.jpg
https://cdn.ncbi.nlm.nih.gov/pmc/blobs/d605/7996210/d91ade4d2589/fncom-15-646125-g0007.jpg

相似文献

1
Spiking Neural Network (SNN) With Memristor Synapses Having Non-linear Weight Update.具有非线性权重更新的忆阻器突触的脉冲神经网络（SNN）
Front Comput Neurosci. 2021 Mar 11;15:646125. doi: 10.3389/fncom.2021.646125. eCollection 2021.
2
On-Chip Training Spiking Neural Networks Using Approximated Backpropagation With Analog Synaptic Devices.使用带有模拟突触器件的近似反向传播的片上训练脉冲神经网络。
Front Neurosci. 2020 Jul 7;14:423. doi: 10.3389/fnins.2020.00423. eCollection 2020.
3
Non-linear Memristive Synaptic Dynamics for Efficient Unsupervised Learning in Spiking Neural Networks.用于脉冲神经网络中高效无监督学习的非线性忆阻突触动力学
Front Neurosci. 2021 Feb 1;15:580909. doi: 10.3389/fnins.2021.580909. eCollection 2021.
4
Mapping the BCPNN Learning Rule to a Memristor Model.将BCPNN学习规则映射到忆阻器模型。
Front Neurosci. 2021 Dec 9;15:750458. doi: 10.3389/fnins.2021.750458. eCollection 2021.
5
Filamentary and Interface-Type Memristors Based on Tantalum Oxide for Energy-Efficient Neuromorphic Hardware.基于氧化钽的丝状和界面型忆阻器用于节能神经形态硬件。
ACS Appl Mater Interfaces. 2022 Oct 5;14(39):44561-44571. doi: 10.1021/acsami.2c12296. Epub 2022 Sep 26.
6
Unsupervised learning in hexagonal boron nitride memristor-based spiking neural networks.基于六方氮化硼忆阻器的脉冲神经网络中的无监督学习。
Nanotechnology. 2023 Aug 17;34(44). doi: 10.1088/1361-6528/acebf5.
7
Hybrid memristor-CMOS neurons for in-situ learning in fully hardware memristive spiking neural networks.用于全硬件忆阻尖峰神经网络原位学习的混合忆阻器-互补金属氧化物半导体神经元。
Sci Bull (Beijing). 2021 Aug 30;66(16):1624-1633. doi: 10.1016/j.scib.2021.04.014. Epub 2021 Apr 17.
8
Hybrid CMOS-Memristor synapse circuits for implementing Ca ion-based plasticity model.用于实现基于钙离子可塑性模型的混合互补金属氧化物半导体-忆阻器突触电路。
Sci Rep. 2024 Aug 2;14(1):17915. doi: 10.1038/s41598-024-68359-x.
9
Optimal Mapping of Spiking Neural Network to Neuromorphic Hardware for Edge-AI.用于边缘 AI 的尖峰神经网络到神经形态硬件的最优映射。
Sensors (Basel). 2022 Sep 24;22(19):7248. doi: 10.3390/s22197248.
10
Supervised Learning in All FeFET-Based Spiking Neural Network: Opportunities and Challenges.基于全铁电场效应晶体管的脉冲神经网络中的监督学习：机遇与挑战。
Front Neurosci. 2020 Jun 24;14:634. doi: 10.3389/fnins.2020.00634. eCollection 2020.

引用本文的文献

1
A Study on the Synaptic Behavior of Al/ZrO/TiO/Al Electronic Bipolar Resistance Switching Memristor.Al/ZrO/TiO/Al 电子双极电阻开关忆阻器的突触行为研究
ACS Appl Mater Interfaces. 2025 Aug 20;17(33):47207-47219. doi: 10.1021/acsami.5c09911. Epub 2025 Aug 8.
2
Defect-Tolerant Memristor Crossbar Circuits for Local Learning Neural Networks.用于局部学习神经网络的容错忆阻器交叉开关电路
Nanomaterials (Basel). 2025 Jan 28;15(3):213. doi: 10.3390/nano15030213.
3
Memristive leaky integrate-and-fire neuron and learnable straight-through estimator in spiking neural networks.

本文引用的文献

1
Rapid online learning and robust recall in a neuromorphic olfactory circuit.神经形态嗅觉回路中的快速在线学习与稳健记忆
Nat Mach Intell. 2020 Mar;2(3):181-191. doi: 10.1038/s42256-020-0159-4. Epub 2020 Mar 16.
2
Fully hardware-implemented memristor convolutional neural network.全硬件实现的忆阻器卷积神经网络。
Nature. 2020 Jan;577(7792):641-646. doi: 10.1038/s41586-020-1942-4. Epub 2020 Jan 29.
3
Flexible Piezoelectric Acoustic Sensors and Machine Learning for Speech Processing.用于语音处理的柔性压电声学传感器和机器学习。
忆阻型漏电积分发放神经元与脉冲神经网络中的可学习直通估计器
Cogn Neurodyn. 2024 Oct;18(5):3075-3091. doi: 10.1007/s11571-024-10133-w. Epub 2024 Jun 20.
4
Bi-sigmoid spike-timing dependent plasticity learning rule for magnetic tunnel junction-based SNN.基于磁性隧道结的脉冲神经网络的双S型脉冲时间依赖可塑性学习规则
Front Neurosci. 2024 May 15;18:1387339. doi: 10.3389/fnins.2024.1387339. eCollection 2024.
5
Memristor Crossbar Circuits Implementing Equilibrium Propagation for On-Device Learning.用于片上学习的实现平衡传播的忆阻交叉开关电路。
Micromachines (Basel). 2023 Jul 3;14(7):1367. doi: 10.3390/mi14071367.
6
Dynamical memristive neural networks and associative self-learning architectures using biomimetic devices.使用仿生器件的动态忆阻神经网络及联想自学习架构
Front Neurosci. 2023 Apr 20;17:1153183. doi: 10.3389/fnins.2023.1153183. eCollection 2023.
7
Convergence of Artificial Intelligence and Neuroscience towards the Diagnosis of Neurological Disorders-A Scoping Review.人工智能与神经科学在神经紊乱诊断中的交汇：综述
Sensors (Basel). 2023 Mar 13;23(6):3062. doi: 10.3390/s23063062.
8
Impact of Asymmetric Weight Update on Neural Network Training With Tiki-Taka Algorithm.非对称权重更新对基于蒂基-塔卡算法的神经网络训练的影响。
Front Neurosci. 2022 Jan 6;15:767953. doi: 10.3389/fnins.2021.767953. eCollection 2021.
Adv Mater. 2020 Sep;32(35):e1904020. doi: 10.1002/adma.201904020. Epub 2019 Oct 16.
4
Improving linearity by introducing Al in HfO as a memristor synapse device.通过在作为忆阻器突触器件的氧化铪中引入铝来提高线性度。
Nanotechnology. 2019 Nov 1;30(44):445205. doi: 10.1088/1361-6528/ab3480.
5
Unsupervised Online Learning With Multiple Postsynaptic Neurons Based on Spike-Timing-Dependent Plasticity Using a Thin-Film Transistor-Type NOR Flash Memory Array.基于薄膜晶体管型 NOR 闪存阵列的基于尖峰时间依赖性可塑性的具有多个突触后神经元的无监督在线学习。
J Nanosci Nanotechnol. 2019 Oct 1;19(10):6050-6054. doi: 10.1166/jnn.2019.17025.
6
Development of a Recognition System for Spraying Areas from Unmanned Aerial Vehicles Using a Machine Learning Approach.利用机器学习方法开发无人飞行器喷洒区域识别系统。
Sensors (Basel). 2019 Jan 14;19(2):313. doi: 10.3390/s19020313.
7
Extended memory lifetime in spiking neural networks employing memristive synapses with nonlinear conductance dynamics.利用具有非线性电导动力学的忆阻突触提高尖峰神经网络的存储时间。
Nanotechnology. 2019 Jan 1;30(1):015102. doi: 10.1088/1361-6528/aae81c.
8
Efficient and self-adaptive in-situ learning in multilayer memristor neural networks.多层忆阻器神经网络中的高效自适应原位学习。
Nat Commun. 2018 Jun 19;9(1):2385. doi: 10.1038/s41467-018-04484-2.
9
Evidence of soft bound behaviour in analogue memristive devices for neuromorphic computing.用于神经形态计算的模拟忆阻器件中的软束缚行为的证据。
Sci Rep. 2018 May 8;8(1):7178. doi: 10.1038/s41598-018-25376-x.
10
Deep Learning with Convolutional Neural Networks Applied to Electromyography Data: A Resource for the Classification of Movements for Prosthetic Hands.基于卷积神经网络的深度学习应用于肌电图数据：一种用于假手运动分类的资源。
Front Neurorobot. 2016 Sep 7;10:9. doi: 10.3389/fnbot.2016.00009. eCollection 2016.