一个基于生物免疫的神经原型,用于几次射击异常检测,并嵌入字符
Zhongjing Ma1, Zhan Chen1, Xiaochen Zheng2
1School of Automation, Beijing Institute of Technology, Beijing 100081, China.
Cyborg and bionic systems (Washington, D.C.)
|January 18, 2024
概括
这项研究引入了一种新的异常检测网络,其灵感来源于对文本数据的生物免疫力. 该方法提高了几次射击检测的准确性和回忆力,即使有有限的标记数据.
科学领域:
- 计算机科学 计算机科学
- 人工智能的人工智能
- 网络安全 网络安全
背景情况:
- 异常检测对于识别入侵和设备故障等各种问题至关重要.
- 在实际场景中,有限的注释数据和可信的标签阻碍了检测性能.
研究的目的:
- 为文本数据提出了一种新的几次射击异常检测网络,灵感来自生物免疫系统.
- 为应对基于文本的系统中低资源异常检测的挑战.
主要方法:
- 一个字符级别表示提取和Word2Vec嵌入方法.
- 一个元学习阶段,使用一个带有编码器,路由和关系模块的动态原型.
- 一个动态路由算法来分配权重,以支持改善原型生成的样本集.
主要成果:
- 拟议的异常检测原型优于最先进的几次射击技术,达到1.3%-4.48%的更高精度和0.18%-4.55%的更高回忆率.
- 在显著减少的训练样本 (5-10) 中,有效的检测得到了维持.
- 废弃研究证实了动态路由算法对更准确原型的贡献.
结论:
- 生物启发的异常检测网络在短时间的文本异常检测中提供了卓越的性能.
- 动态路由机制是创建强大的和准确的异常检测原型的关键.
- 这种方法有效地减轻了在异常检测任务中有限的标记数据的影响.
相关概念视频
Cells of the Adaptive Immune Response
The T and B lymphocytes of the adaptive immune system develop from common lymphoid progenitor cells in the bone marrow. These progenitors give rise to precursors that eventually develop into both T and B lymphocytes. As these precursors mature, they gain the ability to detect and respond to foreign antigens in the body, a process known as immunocompetence. Additionally, these precursors acquire self-tolerance, a process that ensures they do not react to self-antigens. This intricate system...
Immunological Memory
Immunological memory, a pivotal pillar of the adaptive immune system, is responsible for the body's ability to remember and respond more swiftly and effectively to previously encountered pathogens. This remarkable feature is what makes vaccines so effective in preventing diseases.
What is Immunological Memory?
Immunological memory is an integral function of the immune system that allows it to recognize and react more rapidly and effectively to pathogens previously encountered. This feature is...
What is Immunological Memory?
Immunological memory is an integral function of the immune system that allows it to recognize and react more rapidly and effectively to pathogens previously encountered. This feature is...


