深度搜索网络深度搜索网络
IEEE transactions on pattern analysis and machine intelligence
|September 18, 2025
概括
本研究引入了一种新的查找操作,以取代卷积神经网络 (CNN) 中的计算密集的乘法,提高移动设备的效率. 查找网络实现更快的推断和更低的能源使用,同时在图像和点云任务中保持高性能.
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:
- 卷积神经网络 (CNN) 由于诸如乘法之类的运算而具有计算密集性.
- 高计算复杂性阻碍了CNN在资源有限的边缘设备上部署.
- 在CNN中的乘法操作消耗大量能量,并增加推理时间.
研究的目的:
- 引入一种通用且高效的查找操作,作为神经网络的基本构建块.
- 用高效的查找操作取代计算上昂贵的乘法运算.
- 为了实现搜索操作的端到端优化,以提高网络性能.
主要方法:
- 开发了一个可差异化的查找表结构,用于端到端的培训.
- 拟议的培训策略,以确保查找网络的融合.
- 在CNN中用查找操作取代乘法运算,以创建查找网络.
- 应用查找网络用于图像分类,图像超分辨率和点云分类.
主要成果:
- 查找网络显示了能源消耗和推断速度的显著改善.
- 与传统的CNN相比,拟议的查找操作保持了竞争性性能.
- 在各种任务中取得了最先进的结果,包括分类和回归.
- 在不同的数据类型中验证了有效性,例如图像和点云.
结论:
- 查找操作为CNN中的乘法提供了可行和高效的替代方案.
- 查找网络适合在资源受限的边缘设备上部署.
- 该方法为开发更有效的深度学习模型提供了一个有希望的方向.
相关概念视频
Retrieval
410
Retrieval is the process of getting information out of memory storage and back into conscious awareness. This ability is essential for daily tasks like brushing hair and teeth, driving to work, and performing job duties. Retrieval occurs in three ways: recall, recognition, and relearning.
Recall involves accessing information without cues, such as during an essay test, where individuals must retrieve facts and concepts from memory unaided. Another example is remembering the name of a colleague...
Recall involves accessing information without cues, such as during an essay test, where individuals must retrieve facts and concepts from memory unaided. Another example is remembering the name of a colleague...
410
Network Function of a Circuit
652
Frequency response analysis in electrical circuits provides vital insights into a circuit's behavior as the frequency of the input signal changes. The transfer function, a mathematical tool, is instrumental in understanding this behavior. It defines the relationship between phasor output and input and comes in four types: voltage gain, current gain, transfer impedance, and transfer admittance. The critical components of the transfer function are the poles and zeros.
652
Protein Networks
2.8K
2.8K
Protein Networks
4.5K
An organism can have thousands of different proteins, and these proteins must cooperate to ensure the health of an organism. Proteins bind to other proteins and form complexes to carry out their functions. Many proteins interact with multiple other proteins creating a complex network of protein interactions.
These interactions can be represented through maps depicting protein-protein interaction networks, represented as nodes and edges. Nodes are circles that are representative of a protein,...
These interactions can be represented through maps depicting protein-protein interaction networks, represented as nodes and edges. Nodes are circles that are representative of a protein,...
4.5K
ER Retrieval Pathway
4.7K
In the secretory pathway, vesicles transport proteins from one cellular compartment to another in forward transport to deliver the protein to its correct location. Occasionally, misfolded proteins and incorrect proteins escape their original compartments, and a retrieval pathway is used to return the escaped proteins to their original compartment.
The ER uses many checkpoints to prevent the entry of incorrectly folded or a resident protein as cargo onto a transport vesicle. These mechanisms...
The ER uses many checkpoints to prevent the entry of incorrectly folded or a resident protein as cargo onto a transport vesicle. These mechanisms...
4.7K
Storage
364
A schema is a mental framework that helps individuals organize and interpret information. Schemata, formed from previous experiences, influence how we process new information: how we encode it, the inferences we make, and how we retrieve it. For instance, a schema for what a typical classroom looks like might include desks, a teacher's desk, a whiteboard, and students in such an environment. This expectation helps us quickly understand and navigate new classrooms without needing to analyze...
364


