SIA OpenIR  > 机器人学研究室
基于自适应编码的脉冲神经网络
Alternative TitleSelf-adaptive coding for spiking neural network
张驰1,2,3; 唐凤珍1,2
Department机器人学研究室
Source Publication计算机应用研究
ISSN1001-3695
2021
Pages1-6
Contribution Rank1
Funding Organization国家重点研发计划资助项目(2020YFB13400) ; 国家自然科学基金资助项目(61803369)
Keyword脉冲神经网络 自适应编码 替代梯度反向传播 LIF神经元模型
Abstract

脉冲神经网络(SNN)采用脉冲序列表征和传递信息,与传统人工神经网络相比,更具有生物可解释性。但典型SNN的特征提取能力受到其结构限制,对于图像数据等多分类任务的识别准确率不高,不能与卷积神经网络(CNN)相媲美。针对该问题,提出了一种新型的自适应编码脉冲神经网络(SCSNN),将CNN的特征提取能力和SNN的生物可解释性结合起来,采用生物神经元动态脉冲触发特性构建网络结构,并设计了一种新的替代梯度反向传播方法直接训练网络参数。所提出的SCSNN网络分别在MNIST数据集和Fashion-MNIST数据集做了验证,取得较好的识别结果,在MNIST数据集上准确率达到了99.62%,在Fashion-MNIST数据集上准确率达到了93.52%,验证了本模型的有效性。

Other Abstract

Spiking neural networks (SNN) , using spikes to represent and convey information, are more biologically plausible than traditional artificial neural networks. However, a classical shallow SNN has limited feature exraction ability due to the shallow network structure, leading to inferior classification performance to convolutional neural networks (CNN) especially on multi-class classification tasks such as object categorization. In this paper, inspired by the powerful convolutional structure of CNN, a self-adaptive coding spiking neural network (SCSNN) is proposed. By exploiting convolutional structures and dynamic impulse triggerred property of biological neurons, the proposed SCSNN organizes integrate-and-firing models in a convolutional fashion, and trained by a new surrogate gradient back-propagation algorithm directly. The proposed SCSNN network was validated on the MNIST dataset and the Fashion-MNIST dataset, respectively, obtaining superior performance to state-or-the-art SNN networks on both datasets. The classification accuracy on the MNIST dataset reaches 99. 62%, comparable to the performance of exsiting SNN networks. The classification accuracy on the Fashion-MNIST dataset reaches 93. 52%, significantly better than exsiting SNN networks, confirming the effectiveness of the proposed model.

Language中文
Document Type期刊论文
Identifierhttp://ir.sia.cn/handle/173321/29539
Collection机器人学研究室
Corresponding Author唐凤珍
Affiliation1.中国科学院沈阳自动化研究所机器人学国家重点实验室
2.中国科学院机器人与智能制造创新研究院
3.中国科学院大学
Recommended Citation
GB/T 7714
张驰,唐凤珍. 基于自适应编码的脉冲神经网络[J]. 计算机应用研究,2021:1-6.
APA 张驰,&唐凤珍.(2021).基于自适应编码的脉冲神经网络.计算机应用研究,1-6.
MLA 张驰,et al."基于自适应编码的脉冲神经网络".计算机应用研究 (2021):1-6.
Files in This Item:
File Name/Size DocType Version Access License
基于自适应编码的脉冲神经网络.pdf(903KB)期刊论文出版稿开放获取CC BY-NC-SAView Application Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[张驰]'s Articles
[唐凤珍]'s Articles
Baidu academic
Similar articles in Baidu academic
[张驰]'s Articles
[唐凤珍]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[张驰]'s Articles
[唐凤珍]'s Articles
Terms of Use
No data!
Social Bookmark/Share
File name: 基于自适应编码的脉冲神经网络.pdf
Format: Adobe PDF
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.