电子学报 ›› 2022, Vol. 50 ›› Issue (10): 2361-2371.DOI: 10.12263/DZXB.20210443

• 学术论文 • 上一篇    下一篇

基于自监督学习的去流行度偏差推荐方法

张帅1,2, 高旻1,2(), 文俊浩1,2, 熊庆宇1,2, 唐旭2   

  1. 1.信息物理社会可信服务计算教育部重点实验室(重庆大学),重庆 400044
    2.重庆大学大数据与软件学院,重庆 400044
  • 收稿日期:2021-04-07 修回日期:2021-09-02 出版日期:2022-10-25
    • 通讯作者:
    • 高旻
    • 作者简介:
    • 张帅 男,1997年生.现为重庆大学大数据与软件学院在读硕士研究生.主要研究方向为机器学习、推荐系统等.E-mail: zhangshuai@cqu.edu.cn
      高旻(通讯作者) 女,1980年生.工学博士.重庆大学大数据与软件学院副教授、硕士生导师.主要研究方向为推荐系统、异常检测、社会 媒体挖掘.
      文俊浩 男,1969年生.工学博士.重庆大学软件学院教授.主要研究方向为计算智能与推荐系统.E-mail: jhwen@cqu.edu.cn
      熊庆宇 男,1965年生.工学博士.重庆大学大数据与软件学院教授、博士生导师.主要研究方向为智能控制、传感器网络和信息系统.E-mail: xiong03@cqu.edu.cn
      唐旭 男,1983年生.现为重庆大学机械学院在读博士研究生.主要研究方向为智能制造与推荐系统.E-mail: tangxu@cqu.edu.cn
    • 基金资助:
    • 国家自然科学基金 (72161005); 重庆市自然科学基金 (cstc2020jcyj-msxmX0690); 重庆大学中央高校基本科研业务费项目 (2020CDJ-LHZZ-039); 重庆市技术创新与应用发展专项重点项目 (cstc2019jscx-fxydX0012); 重庆市留学人员创业创新支持计划 (cx2020097)

Self-Supervised Learning for Alleviating Popularity Bias in Recommender Systems

ZHANG Shuai1,2, GAO Min1,2(), WEN Jun-hao1,2, XIONG Qing-yu1,2, TANG Xu2   

  1. 1.Key Laboratory of Dependable Service Computing in Cyber Physical Society(Chongqing University),Ministry of Education,Chongqing 400044,China
    2.School of Big Data & Software Engineering,Chongqing University,Chongqing 400044,China
  • Received:2021-04-07 Revised:2021-09-02 Online:2022-10-25 Published:2022-10-11
    • Corresponding author:
    • GAO Min

摘要:

近年来,随着推荐系统研究的不断深入,推荐系统的公平性受到越来越多关注. 流行度偏差也即流行的物品比非流行的物品更容易被推荐,是影响其公平性的重要因素之一. 流行度偏差对推荐系统的各利益相关者都有严重的影响,引起研究者的广泛关注. 相关研究主要通过推荐结果重排或学习过程中融合正则化项提升非流行物品的曝光率,而非流行物品的交互数据极度稀疏成为研究的瓶颈. 针对此问题,本文提出基于自监督学习的去流行度偏差推荐方法,解决两个难点:(1)准确学习交互数据极度稀疏的非流行物品的表征;(2)提升非流行物品曝光率的同时,兼顾不同用户对流行和非流行物品的偏好. 具体地,从用户的角度,提出流行物品和非流行物品双视图的用户偏好学习方法,准确学习用户对流行和非流行物品的真实偏好;从物品的角度,采用自监督学习,利用互信息最大化捕获非流行物品与流行物品间的潜在关系,辅助提升非流行物品嵌入学习的准确性. 最后,设计用户流行度偏好一致性、资格公平性等指标,并通过三个公开数据集的大量实验说明了本文方法在提升推荐性能的同时,能有效缓解流行度偏差问题并具有较强的通用性.

关键词: 推荐系统, 协同过滤, 公平性, 流行度偏差, 自监督学习

Abstract:

In recent years, with the development of recommender system, more and more attention has been paid to the fairness of recommender system. Popularity bias mens that popular items are more likely to be recommended than unpopular items, which is one of the important factors affecting its fairness. Popularity bias has a serious impact on all stakeholders in the recommendation system, which has aroused widespread concern of researchers. Related research mainly improves the exposure rate of unpopular items by the rearrangement of recommended results or the integration of regularization items in the learning process, but the extremely sparse interaction data of unpopular items becomes the bottleneck of research. To solve this problem, this paper proposes self-supervised learning for alleviating popularity bias (SSLAB) to deal with two difficulties: (1) accurately learning the representation of unpopular items with extremely sparse interactive data; (2) improving the exposure rate of unpopular items while taking into account the preferences of different users for popular and unpopular items. Specifically, from the perspective of users, this paper proposes a dual view user preference learning method for popular and unpopular items to accurately learn users’ real preferences for popular and unpopular items; from the perspective of items, self-supervised learning is used to capture the potential relationship between unpopular items and popular items by maximizing mutual information to help improve the accuracy of unpopular items embedded learning. Finally, metrics such as user popularity preference consistency and qualification fairness are designed, and a large number of experiments on three open data sets show that the proposed method can effectively alleviate the popularity bias and has a strong scalability while improving the recommendation performance.

Key words: recommender system, collaborative filtering, fairness, popularity bias, self-supervised learning

中图分类号: