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

Neuroplasticity01:01

Neuroplasticity

361
Neuroplasticity reflects the brain's remarkable capacity to adapt and evolve, responding dynamically to learning, experiences, or injury by reorganizing its neural circuitry. This reorganization involves creating new neural connections and refining old ones through a series of biological processes that contribute to the brain's lifelong development and adaptability.
361

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 journal

Impact of room configuration on energy use intensity in high-density pilgrim accommodations: a case study in Makkah, Saudi Arabia.

Scientific reports·2026
Same journal

Retraction Note: Improving the power production efficiency of microbial fuel cell by using biosynthesized polyanaline coated Fe<sub>3</sub>O<sub>4</sub> as pencil graphite anode modifier.

Scientific reports·2026
Same journal

Stylopine as multi-target anti-Alzheimer agent.

Scientific reports·2026
Same journal

Principal component analysis of morphological features of the aortic arch for predicting the risk for acute type B aortic dissection.

Scientific reports·2026
Same journal

An explainable AfroXLMR approach for multi-label emotion classification of Amharic social media text with dataset release.

Scientific reports·2026
Same journal

Bridging in vitro bioactivities and in silico insights of Rubia cordifolia L. (Rubiaceae) leaf extracts for therapeutic applications.

Scientific reports·2026

相关实验视频

Updated: Jul 4, 2025

Deep Neural Networks for Image-Based Dietary Assessment
13:19

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

9.2K

资源有限的神经网络培训

Mariusz Pietrołaj1, Marek Blok2

  • 1Faculty of Electronics, Telecommunications, and Informatics, Gdansk University of Technology, Gdańsk, Poland. mariusz.pietrolaj@pg.edu.pl.

Scientific reports
|January 29, 2024
PubMed
概括
此摘要是机器生成的。

这项研究完善了边缘AI的神经网络参数限制. 使用非对称指数方法,研究人员训练了具有8位浮点精度的模型,保持与32位系统可比的性能.

更多相关视频

Closed-loop Neuro-robotic Experiments to Test Computational Properties of Neuronal Networks
11:18

Closed-loop Neuro-robotic Experiments to Test Computational Properties of Neuronal Networks

Published on: March 2, 2015

10.3K
Training Dogs for Awake, Unrestrained Functional Magnetic Resonance Imaging
07:59

Training Dogs for Awake, Unrestrained Functional Magnetic Resonance Imaging

Published on: October 13, 2019

7.6K

相关实验视频

Last Updated: Jul 4, 2025

Deep Neural Networks for Image-Based Dietary Assessment
13:19

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

9.2K
Closed-loop Neuro-robotic Experiments to Test Computational Properties of Neuronal Networks
11:18

Closed-loop Neuro-robotic Experiments to Test Computational Properties of Neuronal Networks

Published on: March 2, 2015

10.3K
Training Dogs for Awake, Unrestrained Functional Magnetic Resonance Imaging
07:59

Training Dogs for Awake, Unrestrained Functional Magnetic Resonance Imaging

Published on: October 13, 2019

7.6K

科学领域:

  • 人工智能的人工智能
  • 计算机科学 计算机科学
  • 机器学习 机器学习

背景情况:

  • 现代人工智能应用正在转向资源受限的边缘设备.
  • 这就需要有效的神经网络训练方法,以减少计算和功率需求.
  • 现有的参数位数减少技术主要集中在推断上,而不是培训.

研究的目的:

  • 开发和评估一种用于在训练期间限制神经网络参数的新方法.
  • 研究浮点变量表示和圆的先进技术.
  • 为了实现有效的神经网络训练,而不会显著降低性能.

主要方法:

  • 实施一个不对称的指数方法来限制参数.
  • 探索新的浮点变量表示和圆策略.
  • 使用指数偏移来进行浮点精度调整,而不会增加位数.

主要成果:

  • 成功训练了LeNet,AlexNet和ResNet-18卷积神经网络,使用自定义的8位浮点表示.
  • 与基线32位浮点训练相比,在性能方面实现了最小的或没有退化.
  • 证明了资源高效的神经网络培训的可行性.

结论:

  • 非对称指数方法提供了一种有效的方法来减少神经网络训练中的资源需求.
  • 8位浮点精度可用于训练复杂模型,精度可与32位相比较.
  • 这项研究使得更具成本效益的AI模型开发和在个人设备上更广泛的AI应用成为可能.