From Social to Individuals: A Parsimonious Path of Multi-Level Models for Crowdsourced Preference Aggregation

doi:10.1109/TPAMI.2018.2817205

CORC > 计算技术研究所 > 中国科学院计算技术研究所 > 中国科学院计算技术研究所期刊论文 > 英文

	From Social to Individuals: A Parsimonious Path of Multi-Level Models for Crowdsourced Preference Aggregation
	Yao, Yuan 1,2; Huang, Qingming 3,4; Cao, Xiaochun 7; Xiong, Jiechao 5,6; Xu, Qianqian 7,8
刊名	IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
	2019-04-01
卷号	41 期号:4 页码:844-856
关键词	Preference aggregation HodgeRank mixed-effects models linearized bregman iterations personalized ranking position bias
ISSN号	0162-8828
DOI	10.1109/TPAMI.2018.2817205
英文摘要	In crowdsourced preference aggregation, it is often assumed that all the annotators are subject to a common preference or social utility function which generates their comparison behaviors in experiments. However, in reality, annotators are subject to variations due to multi-criteria, abnormal, or a mixture of such behaviors. In this paper, we propose a parsimonious mixed-effects model, which takes into account both the fixed effect that the majority of annotators follows a common linear utility model, and the random effect that some annotators might deviate from the common significantly and exhibit strongly personalized preferences. The key algorithm in this paper establishes a dynamic path from the social utility to individual variations, with different levels of sparsity on personalization. The algorithm is based on the Linearized Bregman Iterations, which leads to easy parallel implementations to meet the need of large-scale data analysis. In this unified framework, three kinds of random utility models are presented, including the basic linear model with L-2 loss, Bradley-Terry model, and Thurstone-Mosteller model. The validity of these multi-level models are supported by experiments with both simulated and real-world datasets, which shows that the parsimonious multi-level models exhibit improvements in both interpretability and predictive precision compared with traditional HodgeRank.
资助项目	National Key Research and Development Plan[2016YFB0800403] ; National Natural Science Foundation of China[61672514] ; National Natural Science Foundation of China[61390514] ; National Natural Science Foundation of China[61572042] ; National Natural Science Foundation of China[U1636214] ; National Natural Science Foundation of China[61650202] ; National Natural Science Foundation of China[61332016] ; National Natural Science Foundation of China[61620106009] ; National Natural Science Foundation of China[61370004] ; National Natural Science Foundation of China[11421110001] ; Beijing Natural Science Foundation[4182079] ; Beijing Natural Science Foundation[4172068] ; Youth Innovation Promotion Association CAS ; CCF-Tencent Open Research Fund ; Key Program of the Chinese Academy of Sciences[QYZDB-SSW-JSC003] ; National Basic Research Program of China (973 Program)[2015CB351800] ; Key Research Program of Frontier Sciences[CAS: QYZDJ-SSW-SYS013] ; Hong Kong Research Grant Council (HKRGC)[16303817] ; National Basic Research Program of China[2015CB85600] ; National Basic Research Program of China[2012CB825501] ; Tencent AI Lab ; Si Family Foundation ; Baidu Big Data Institute ; Microsoft Research-Asia
WOS研究方向	Computer Science ; Engineering
语种	英语
出版者	IEEE COMPUTER SOC
WOS记录号	WOS:000460583500005
内容类型	期刊论文
源URL	[http://119.78.100.204/handle/2XEOYT63/4111]
专题	中国科学院计算技术研究所期刊论文_英文
通讯作者	Yao, Yuan; Huang, Qingming; Cao, Xiaochun
作者单位	1.Hong Kong Univ Sci & Technol, Dept Comp Sci & Engn, Hong Kong, Peoples R China 2.Hong Kong Univ Sci & Technol, Dept Math, Hong Kong, Peoples R China 3.Chinese Acad Sci, Inst Comp Technol, Beijing 100190, Peoples R China 4.Univ Chinese Acad Sci, Huairou 101408, Peoples R China 5.Peking Univ, BICMR LMAM LMEQF LMP, Sch Math Sci, Beijing 100871, Peoples R China 6.Tencent AI Lab, Shenzhen 518057, Peoples R China 7.Chinese Acad Sci, Inst Informat Engn, SKLOIS, Beijing 100093, Peoples R China 8.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	Yao, Yuan,Huang, Qingming,Cao, Xiaochun,et al. From Social to Individuals: A Parsimonious Path of Multi-Level Models for Crowdsourced Preference Aggregation[J]. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE,2019,41(4):844-856.
APA	Yao, Yuan,Huang, Qingming,Cao, Xiaochun,Xiong, Jiechao,&Xu, Qianqian.(2019).From Social to Individuals: A Parsimonious Path of Multi-Level Models for Crowdsourced Preference Aggregation.IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE,41(4),844-856.
MLA	Yao, Yuan,et al."From Social to Individuals: A Parsimonious Path of Multi-Level Models for Crowdsourced Preference Aggregation".IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 41.4(2019):844-856.