Jove
Visualize
联系我们
JoVE
x logofacebook logolinkedin logoyoutube logo
关于 JoVE
概览领导团队博客JoVE 帮助中心
作者
出版流程编辑委员会范围与政策同行评审常见问题投稿
图书馆员
用户评价订阅访问资源图书馆顾问委员会常见问题
研究
JoVE JournalMethods CollectionsJoVE Encyclopedia of Experiments存档
教育
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab Manual教师资源中心教师网站
使用条款与条件
隐私政策
政策

相关概念视频

您也可能阅读

相关文章

通过共同作者、期刊和引用图与本文相关的文章。

排序
Same author

The purine-rich element-binding protein ChPur-α negatively regulates Hsc70 transcription in Crassostrea hongkongensis.

Cell stress & chaperones·2017
Same author

Improved antitumor effect of ionizing radiation in combination with rapamycin for treating nasopharyngeal carcinoma.

Oncology letters·2017
Same author

Roles of Cells from the Arterial Vessel Wall in Atherosclerosis.

Mediators of inflammation·2017
Same author

Metabolic and microbial signatures in rat hepatocellular carcinoma treated with caffeic acid and chlorogenic acid.

Scientific reports·2017
Same author

Arsenic removal in aqueous solution by a novel Fe-Mn modified biochar composite: Characterization and mechanism.

Ecotoxicology and environmental safety·2017
Same author

Antidiabetic activities of polysaccharides separated from Inonotus obliquus via the modulation of oxidative stress in mice with streptozotocin-induced diabetes.

PloS one·2017

相关实验视频

Updated: May 22, 2025

Demonstration of Spin-Multiplexed and Direction-Multiplexed All-Dielectric Visible Metaholograms
08:48

Demonstration of Spin-Multiplexed and Direction-Multiplexed All-Dielectric Visible Metaholograms

Published on: September 25, 2020

5.7K

基于双面元表面的可翻转多任务衍射神经网络.

He Ren, Shuai Zhou, Yuxiang Feng

    Optics letters
    |March 14, 2025
    PubMed
    概括
    此摘要是机器生成的。

    可翻转衍射神经网络 (F-DNN) 通过使用双面衍射层来快速切换任务,使多任务实现. 这种物理计算方法为人工智能系统提供了高速,低功耗和可扩展性.

    更多相关视频

    Demonstration of Equal-Intensity Beam Generation by Dielectric Metasurfaces
    09:33

    Demonstration of Equal-Intensity Beam Generation by Dielectric Metasurfaces

    Published on: June 7, 2019

    6.2K
    Revised and Neuroimaging-Compatible Versions of the Dual Task Screen
    07:52

    Revised and Neuroimaging-Compatible Versions of the Dual Task Screen

    Published on: October 5, 2020

    3.5K

    相关实验视频

    Last Updated: May 22, 2025

    Demonstration of Spin-Multiplexed and Direction-Multiplexed All-Dielectric Visible Metaholograms
    08:48

    Demonstration of Spin-Multiplexed and Direction-Multiplexed All-Dielectric Visible Metaholograms

    Published on: September 25, 2020

    5.7K
    Demonstration of Equal-Intensity Beam Generation by Dielectric Metasurfaces
    09:33

    Demonstration of Equal-Intensity Beam Generation by Dielectric Metasurfaces

    Published on: June 7, 2019

    6.2K
    Revised and Neuroimaging-Compatible Versions of the Dual Task Screen
    07:52

    Revised and Neuroimaging-Compatible Versions of the Dual Task Screen

    Published on: October 5, 2020

    3.5K

    科学领域:

    • 光学和光子学 在光学和光子学.
    • 人工智能的人工智能
    • 物理计算 物理计算

    背景情况:

    • 分散神经网络 (DNN) 提供高速,低功耗的计算,但由于不可重新配置的层,难以实现多任务处理.
    • 现有的多任务DNN的方法,如层或光源更换,是不切实际的.
    • 调整问题阻碍了传统多任务DNN架构的实施.

    研究的目的:

    • 介绍一个新的可翻转衍射神经网络 (F-DNN) 架构.
    • 解决传统DNN在高效执行多个任务方面的局限性.
    • 为多任务物理计算提供可扩展和实用的解决方案.

    主要方法:

    • 设计了一个集成的衍射层,在基板的两侧处理.
    • 实现了"翻转"机制,以便在不同的衍射模式之间快速切换.
    • 利用基于分类的模拟来评估F-DNN的性能.

    主要成果:

    • 通过层翻转,F-DNN架构成功实现了快速的任务切换.
    • 克服了与DNN中层替换相关的实际对齐挑战.
    • 与传统的多任务DNN相比,模拟结果显示出更高的性能和可扩展性.

    结论:

    • 可翻转的衍射神经网络为多任务DNN提供了可行的解决方案.
    • 这种方法提高了物理AI系统的速度,功率效率和多任务能力.
    • F-DNN为开发先进,多功能的人工智能硬件提供了新的途径.