Learning to Sketch: A Neural Approach to Item Frequency Estimation in Streaming Data

doi:10.1109/TPAMI.2024.3388589

	Learning to Sketch: A Neural Approach to Item Frequency Estimation in Streaming Data
	Cao, Yukun 1,2; Feng, Yuan 1,2; Wang, Hairu 1,2; Xie, Xike 2,3; Zhou, S. Kevin 4,5,6
	2024-11-01
发表期刊	IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
ISSN	0162-8828
卷号	46 期号:11 页码:7136-7153
摘要	Recently, there has been a trend of designing neural data structures to go beyond handcrafted data structures by leveraging patterns of data distributions for better accuracy and adaptivity. Sketches are widely used data structures in real-time web analysis, network monitoring, and self-driving to estimate item frequencies of data streams within limited space. However, existing sketches have not fully exploited the patterns of the data stream distributions, making it challenging to tightly couple them with neural networks that excel at memorizing pattern information. Starting from the premise, we envision a pure neural data structure as a base sketch, which we term the meta-sketch, to reinvent the base structure of conventional sketches. The meta-sketch learns basic sketching abilities from meta-tasks constituted with synthetic datasets following Zipf distributions in the pre-training phase and can be quickly adapted to real (skewed) distributions in the adaption phase. The meta-sketch not only surpasses its competitors in sketching conventional data streams but also holds good potential in supporting more complex streaming data, such as multimedia and graph stream scenarios. Extensive experiments demonstrate the superiority of the meta-sketch and offer insights into its working mechanism.
关键词	sketches meta-learning Neural data structure memory-augmented neural networks memory-augmented neural networks meta-learning memory-augmented neural networks
DOI	10.1109/TPAMI.2024.3388589
收录类别	SCI
语种	英语
资助项目	National Natural Science Foundation of China[61772492] ; National Natural Science Foundation of China[62072428] ; National Natural Science Foundation of China[62271465] ; Open Fund Project of Guangdong Academy of Medical Sciences, China[YKY-KF202206] ; Suzhou Basic Research Program[SYG202338]
WOS研究方向	Computer Science ; Engineering
WOS类目	Computer Science, Artificial Intelligence ; Engineering, Electrical & Electronic
WOS记录号	WOS:001329047900010
出版者	IEEE COMPUTER SOC
引用统计
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/39528
专题	中国科学院计算技术研究所期刊论文_英文
通讯作者	Xie, Xike; Zhou, S. Kevin
作者单位	1.Univ Sci & Technol China USTC, Sch Comp Sci & Technol, Hefei 230026, Peoples R China 2.Univ Sci & Technol China, Suzhou Inst Adv Res, Ctr Med Imaging Robot Analyt Comp & Learning MIR A, Data Darkness Lab DDL, Hefei 230026, Peoples R China 3.Univ Sci & Technol China, Sch Biomed Engn, Hefei 230026, Peoples R China 4.Univ Sci & Technol China, Sch Biomed Engn, Div Life Sci & Med, Hefei 230026, Peoples R China 5.Univ Sci & Technol China, Suzhou Inst Adv Res, Ctr Med Imaging Robot Analyt Comp & Learning MIRAC, Hefei 230026, Peoples R China 6.Chinese Acad Sci, Inst Comp Technol, Key Lab Intelligent Informat Proc, Beijing 100190, Peoples R China
推荐引用方式 GB/T 7714	Cao, Yukun,Feng, Yuan,Wang, Hairu,et al. Learning to Sketch: A Neural Approach to Item Frequency Estimation in Streaming Data[J]. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE,2024,46(11):7136-7153.
APA	Cao, Yukun,Feng, Yuan,Wang, Hairu,Xie, Xike,&Zhou, S. Kevin.(2024).Learning to Sketch: A Neural Approach to Item Frequency Estimation in Streaming Data.IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE,46(11),7136-7153.
MLA	Cao, Yukun,et al."Learning to Sketch: A Neural Approach to Item Frequency Estimation in Streaming Data".IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 46.11(2024):7136-7153.