打开深度学习盒子

IF 21.2 1区 医学 Q1 NEUROSCIENCES
Luis A. Mejia
{"title":"打开深度学习盒子","authors":"Luis A. Mejia","doi":"10.1038/s41593-025-01938-x","DOIUrl":null,"url":null,"abstract":"","PeriodicalId":19076,"journal":{"name":"Nature neuroscience","volume":"28 4","pages":"709-709"},"PeriodicalIF":21.2000,"publicationDate":"2025-04-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Opening the deep learning box\",\"authors\":\"Luis A. Mejia\",\"doi\":\"10.1038/s41593-025-01938-x\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"\",\"PeriodicalId\":19076,\"journal\":{\"name\":\"Nature neuroscience\",\"volume\":\"28 4\",\"pages\":\"709-709\"},\"PeriodicalIF\":21.2000,\"publicationDate\":\"2025-04-04\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Nature neuroscience\",\"FirstCategoryId\":\"3\",\"ListUrlMain\":\"https://www.nature.com/articles/s41593-025-01938-x\",\"RegionNum\":1,\"RegionCategory\":\"医学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"NEUROSCIENCES\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Nature neuroscience","FirstCategoryId":"3","ListUrlMain":"https://www.nature.com/articles/s41593-025-01938-x","RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"NEUROSCIENCES","Score":null,"Total":0}
引用次数: 0

摘要

基于深度学习的神经数据分析可以提取潜在表征,但由于其“黑箱”性质,往往缺乏可解释性。Tolooshams、Matias等人开发了一种基于深度学习的反卷积分析框架,用于学习局部低秩结构,该框架将算法展开与卷积稀疏编码结合为生成模型。该方法通过模型优化学习核重构神经数据,并将神经活动建模为核与稀疏码之间的卷积组合。重要的是,学习到的核和稀疏代码是可直接解释的,这有助于映射神经元的反应。这个框架被称为反卷积展开神经学习(DUNL),它可以反卷积个体多巴胺神经元奖励反应中的显著性和价值成分。它可以以一种无监督的方式学习核,正如来自体感丘脑的记录所展示的那样。DUNL甚至可以用来模拟自然的、非结构化的实验中的神经反应。使用可解释的深度神经网络对神经活动进行反卷积是基于模型的神经数据分析的重要下一章。原始参考文献:Neuron https://doi.org/10.1016/j.neuron.2025.02.006 (2025)
本文章由计算机程序翻译,如有差异,请以英文原文为准。
Opening the deep learning box
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
Nature neuroscience
Nature neuroscience 医学-神经科学
CiteScore
38.60
自引率
1.20%
发文量
212
审稿时长
1 months
期刊介绍: Nature Neuroscience, a multidisciplinary journal, publishes papers of the utmost quality and significance across all realms of neuroscience. The editors welcome contributions spanning molecular, cellular, systems, and cognitive neuroscience, along with psychophysics, computational modeling, and nervous system disorders. While no area is off-limits, studies offering fundamental insights into nervous system function receive priority. The journal offers high visibility to both readers and authors, fostering interdisciplinary communication and accessibility to a broad audience. It maintains high standards of copy editing and production, rigorous peer review, rapid publication, and operates independently from academic societies and other vested interests. In addition to primary research, Nature Neuroscience features news and views, reviews, editorials, commentaries, perspectives, book reviews, and correspondence, aiming to serve as the voice of the global neuroscience community.
×
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
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
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 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信