利用递归尖峰神经网络从皮层尖峰序列解码手指速度

Tengjun Liu, Julia Gygax, Julian Rossbroich, Yansong Chua, Shaomin Zhang, Friedemann Zenke
{"title":"利用递归尖峰神经网络从皮层尖峰序列解码手指速度","authors":"Tengjun Liu, Julia Gygax, Julian Rossbroich, Yansong Chua, Shaomin Zhang, Friedemann Zenke","doi":"arxiv-2409.01762","DOIUrl":null,"url":null,"abstract":"Invasive cortical brain-machine interfaces (BMIs) can significantly improve\nthe life quality of motor-impaired patients. Nonetheless, externally mounted\npedestals pose an infection risk, which calls for fully implanted systems. Such\nsystems, however, must meet strict latency and energy constraints while\nproviding reliable decoding performance. While recurrent spiking neural\nnetworks (RSNNs) are ideally suited for ultra-low-power, low-latency processing\non neuromorphic hardware, it is unclear whether they meet the above\nrequirements. To address this question, we trained RSNNs to decode finger\nvelocity from cortical spike trains (CSTs) of two macaque monkeys. First, we\nfound that a large RSNN model outperformed existing feedforward spiking neural\nnetworks (SNNs) and artificial neural networks (ANNs) in terms of their\ndecoding accuracy. We next developed a tiny RSNN with a smaller memory\nfootprint, low firing rates, and sparse connectivity. Despite its reduced\ncomputational requirements, the resulting model performed substantially better\nthan existing SNN and ANN decoders. Our results thus demonstrate that RSNNs\noffer competitive CST decoding performance under tight resource constraints and\nare promising candidates for fully implanted ultra-low-power BMIs with the\npotential to revolutionize patient care.","PeriodicalId":501517,"journal":{"name":"arXiv - QuanBio - Neurons and Cognition","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2024-09-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Decoding finger velocity from cortical spike trains with recurrent spiking neural networks\",\"authors\":\"Tengjun Liu, Julia Gygax, Julian Rossbroich, Yansong Chua, Shaomin Zhang, Friedemann Zenke\",\"doi\":\"arxiv-2409.01762\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Invasive cortical brain-machine interfaces (BMIs) can significantly improve\\nthe life quality of motor-impaired patients. Nonetheless, externally mounted\\npedestals pose an infection risk, which calls for fully implanted systems. Such\\nsystems, however, must meet strict latency and energy constraints while\\nproviding reliable decoding performance. While recurrent spiking neural\\nnetworks (RSNNs) are ideally suited for ultra-low-power, low-latency processing\\non neuromorphic hardware, it is unclear whether they meet the above\\nrequirements. To address this question, we trained RSNNs to decode finger\\nvelocity from cortical spike trains (CSTs) of two macaque monkeys. First, we\\nfound that a large RSNN model outperformed existing feedforward spiking neural\\nnetworks (SNNs) and artificial neural networks (ANNs) in terms of their\\ndecoding accuracy. We next developed a tiny RSNN with a smaller memory\\nfootprint, low firing rates, and sparse connectivity. Despite its reduced\\ncomputational requirements, the resulting model performed substantially better\\nthan existing SNN and ANN decoders. Our results thus demonstrate that RSNNs\\noffer competitive CST decoding performance under tight resource constraints and\\nare promising candidates for fully implanted ultra-low-power BMIs with the\\npotential to revolutionize patient care.\",\"PeriodicalId\":501517,\"journal\":{\"name\":\"arXiv - QuanBio - Neurons and Cognition\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-09-03\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - QuanBio - Neurons and Cognition\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2409.01762\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - QuanBio - Neurons and Cognition","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2409.01762","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0

摘要

侵入性皮层脑机接口(BMI)可显著改善运动障碍患者的生活质量。然而,外部安装的基座存在感染风险,因此需要完全植入的系统。然而,这种系统必须满足严格的延迟和能量限制,同时提供可靠的解码性能。虽然递归尖峰神经网络(RSNN)非常适合在神经形态硬件上进行超低功耗、低延迟处理,但目前还不清楚它们是否满足上述要求。为了解决这个问题,我们训练 RSNN 从两只猕猴的皮层尖峰序列 (CST) 中解码手指速度。首先,我们发现大型 RSNN 模型的解码准确性优于现有的前馈尖峰神经网络 (SNN) 和人工神经网络 (ANN)。接下来,我们开发了一种微型 RSNN,它具有较小的内存空间、较低的发射率和稀疏的连接性。尽管计算要求降低了,但由此产生的模型的性能却大大优于现有的 SNN 和 ANN 解码器。因此,我们的研究结果表明,在资源紧张的情况下,RSNN 也能提供具有竞争力的 CST 解码性能,是完全植入式超低功耗 BMI 的理想候选者,有望彻底改变病人护理方式。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
查看原文
分享 分享
微信好友 朋友圈 QQ好友 复制链接
本刊更多论文
Decoding finger velocity from cortical spike trains with recurrent spiking neural networks
Invasive cortical brain-machine interfaces (BMIs) can significantly improve the life quality of motor-impaired patients. Nonetheless, externally mounted pedestals pose an infection risk, which calls for fully implanted systems. Such systems, however, must meet strict latency and energy constraints while providing reliable decoding performance. While recurrent spiking neural networks (RSNNs) are ideally suited for ultra-low-power, low-latency processing on neuromorphic hardware, it is unclear whether they meet the above requirements. To address this question, we trained RSNNs to decode finger velocity from cortical spike trains (CSTs) of two macaque monkeys. First, we found that a large RSNN model outperformed existing feedforward spiking neural networks (SNNs) and artificial neural networks (ANNs) in terms of their decoding accuracy. We next developed a tiny RSNN with a smaller memory footprint, low firing rates, and sparse connectivity. Despite its reduced computational requirements, the resulting model performed substantially better than existing SNN and ANN decoders. Our results thus demonstrate that RSNNs offer competitive CST decoding performance under tight resource constraints and are promising candidates for fully implanted ultra-low-power BMIs with the potential to revolutionize patient care.
求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
自引率
0.00%
发文量
0
期刊最新文献
Early reduced dopaminergic tone mediated by D3 receptor and dopamine transporter in absence epileptogenesis Contrasformer: A Brain Network Contrastive Transformer for Neurodegenerative Condition Identification Identifying Influential nodes in Brain Networks via Self-Supervised Graph-Transformer Contrastive Learning in Memristor-based Neuromorphic Systems Self-Attention Limits Working Memory Capacity of Transformer-Based Models
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
现在去查看 取消
×
提示
确定
0
微信
客服QQ
Book学术公众号 扫码关注我们
反馈
×
意见反馈
请填写您的意见或建议
请填写您的手机或邮箱
已复制链接
已复制链接
快去分享给好友吧!
我知道了
×
扫码分享
扫码分享
Book学术官方微信
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术
文献互助 智能选刊 最新文献 互助须知 联系我们:info@booksci.cn
Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。
Copyright © 2023 Book学术 All rights reserved.
ghs 京公网安备 11010802042870号 京ICP备2023020795号-1