量子化神经网络的垂直分层用于异质推理
IEEE transactions on pattern analysis and machine intelligence
|September 25, 2023
概括
这项研究引入了一个神经网络的新的垂直层重量表示,使单个模型能够支持各种量子化精度. 一种新的定量化意识的训练方法可以通过一次性训练在所有支持的比特宽度上确保高性能.
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:
- 神经网络量化对于高效的推断至关重要,但缺乏跨多种硬件的可扩展性.
- 目前的方法需要每个硬件单独的模型,增加培训和维护成本.
研究的目的:
- 开发一个统一的神经网络表示,用于多个量子化精度.
- 为这种统一代表性提出一个有效的培训方法.
主要方法:
- 引入了垂直层重量表示,从最重要到最不重要的位进行组织.
- 开发了一种新的量子化意识培训 (QAT) 方案,使用级联下降采样和多目标优化.
- 在CIFAR-100和ImageNet数据集上评估性能.
主要成果:
- 垂直层次的表示有效地将多个量子化模型封装成一个.
- 建议的一次QAT方案可以实现各种位宽的高性能.
- 性能与为特定位宽度训练的模型相提并论.
结论:
- 垂直层表示和一次 QAT 方案为高效的神经网络部署提供了可扩展的解决方案.
- 这种方法显著减少了对异质设备的培训和维护开销.
相关概念视频
Neural Circuits
1.3K
Neural circuits and neuronal pools are two of the main structures found in the nervous system. Neural circuits are networks of neurons that work together to carry out a specific task or process. They consist of interconnected neurons and glial cells, which provide structural and metabolic support.
Neuronal pools are collections of nerve cells with similar functions and interact through chemical and electrical signals. These pools include both interneurons (the central neural circuit nodes that...
Neuronal pools are collections of nerve cells with similar functions and interact through chemical and electrical signals. These pools include both interneurons (the central neural circuit nodes that...
1.3K
Multi-input and Multi-variable systems
122
Cruise control systems in cars are designed as multi-input systems to maintain a driver's desired speed while compensating for external disturbances such as changes in terrain. The block diagram for a cruise control system typically includes two main inputs: the desired speed set by the driver and any external disturbances, such as the incline of the road. By adjusting the engine throttle, the system maintains the vehicle's speed as close to the desired value as possible.
In the absence...
In the absence...
122
Multicompartment Models: Overview
170
Multicompartment models are mathematical constructs that depict how drugs are distributed and eliminated within the body. They segment the body into several compartments, symbolizing various physiological or anatomical areas connected through drug transfer processes such as absorption, metabolism, distribution, and elimination.
These models offer a more comprehensive representation of drug behavior in the body than one-compartment models. They accommodate the complexity of drug distribution,...
These models offer a more comprehensive representation of drug behavior in the body than one-compartment models. They accommodate the complexity of drug distribution,...
170
Neural Regulation
39.5K
Digestion begins with a cephalic phase that prepares the digestive system to receive food. When our brain processes visual or olfactory information about food, it triggers impulses in the cranial nerves innervating the salivary glands and stomach to prepare for food.
39.5K
Insensitive Nuclei Enhanced by Polarization Transfer (INEPT)
368
Insensitive Nuclei Enhanced by Polarization Transfer (INEPT) is an advanced Nuclear Magnetic Resonance (NMR) technique specifically designed to detect and enhance the signals of low-abundance nuclei, such as carbon-13 and nitrogen-15, in small molecules. The fundamental principle behind INEPT is the transfer of polarization from a more abundant and highly polarizable nucleus, typically hydrogen-1, to the low-abundance nucleus of interest. This process effectively boosts the NMR signal of the...
368


