通过蒸改善轻量级的AdderNet 从l2到l1-标准
概括
添加器神经网络 (ANN) 使用加法而不是乘法来提高效率. 一种新的规范导向蒸 (NGD) 方法通过从l2规范网络中学习来改进ANN,提高轻量级模型的性能.
科学领域:
- 计算机科学 计算机科学
- 人工智能的人工智能
- 机器学习 机器学习
背景情况:
- 卷积神经网络 (CNN) 依赖于计算上昂贵的乘法运算.
- 添加器神经网络 (ANN) 提供了一个硬件友好的替代方案,通过使用l1-norm.用加法替换乘法.
- 在CNN和ANN之间存在性能差距,特别是在降低参数的情况下,目前的方法无法解决这一问题.
研究的目的:
- 引入一种新的方法,即规范引导蒸 (NGD),以提高l1-规范ANN的性能.
- 为了使l1-规范ANN能够从l2-规范ANN有效地学习,弥合绩效差距.
主要方法:
- 拟议的标准导向蒸 (NGD) 对于l1-标准ANNs.
- 利用l2距离在l2规范ANN中的优越集群性能来实现特征学习.
- 在l2-规范ANN的特点中鼓励了类内集中和类间分散.
- 修改了ANN梯度,以便从l2规范逐步接近l1规范,以实现精确的优化.
主要成果:
- NGD显著提高了轻量级ANN的性能.
- 在CIFAR-100和ImageNet.Net等基准标准上表现出有效性.
- 在CIFAR-100上,通过0.25xGhostNet实现了10.43%的改进.
- 在ImageNet.Net上使用1.0x GhostNet实现了3.1%的改进.
结论:
- NGD是一种有效的方法来增强l1-标准的ANN,特别是在资源有限的应用程序中.
- 提出的技术成功地弥合了基于l1标准和l2标准的网络之间的性能差距.
- NGD提供了一种实用的解决方案,用于对硬件友好的设计进行高效的深度学习推断.
更多相关视频
08:54Determining Four Components in a Lipid Nanoparticle RNA Delivery System by Liquid Chromatography Combined with Evaporative Light Scattering Detector
Published on: May 30, 2025
688
05:16Fabricating Multi-Component Lipid Nanotube Networks Using the Gliding Kinesin Motility Assay
Published on: July 26, 2021
1.7K
相关概念视频
Distillation: Vapor–Liquid Equilibria
2.8K
Distillation is a separation technique that takes advantage of the boiling point properties of disparate elements in a mixture. To perform distillation, we begin by heating a miscible mixture of two liquids with a significant difference in boiling points (at least 20°C). As the solution heats up and reaches the bubble point of the more volatile component, some molecules of the more volatile component transition into the gas phase and travel upward into the condenser, which is a glass tube...
2.8K
Reducing Line Loss
168
In a three-phase circuit, line loss is an indicator of energy dissipated as heat due to the resistance of transmission lines. To address this, incorporating transformers into the system—a step-up transformer at the source and a step-down transformer at the load—is a strategic solution. Two three-phase transformers are introduced to improve this.
With a step-up transformer at the source, the voltage is increased, thereby reducing the current in the transmission lines since power loss...
With a step-up transformer at the source, the voltage is increased, thereby reducing the current in the transmission lines since power loss...
168
