uCTRL: Unbiased Contrastive Representation Learning via Alignment and Uniformity for Collaborative Filtering

被引:3
|
作者
Lee, Jae-woong [1 ]
Park, Seongmin [1 ]
Yoon, Mincheol [1 ]
Lee, Jongwuk [1 ]
机构
[1] Sungkyunkwan Univ, Seoul, South Korea
来源
PROCEEDINGS OF THE 46TH INTERNATIONAL ACM SIGIR CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION RETRIEVAL, SIGIR 2023 | 2023年
关键词
Alignment and uniformity; popularity bias;
D O I
10.1145/3539618.3592076
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Because implicit user feedback for the collaborative filtering (CF) models is biased toward popular items, CF models tend to yield recommendation lists with popularity bias. Previous studies have utilized inverse propensity weighting (IPW) or causal inference to mitigate this problem. However, they solely employ pointwise or pairwise loss functions and neglect to adopt a contrastive loss function for learning meaningful user and item representations. In this paper, we propose Unbiased ConTrastive Representation Learning (uCTRL), optimizing alignment and uniformity functions derived from the InfoNCE loss function for CF models. Specifically, we formulate an unbiased alignment function used in uCTRL. We also devise a novel IPW estimation method that removes the bias of both users and items. Despite its simplicity, uCTRL equipped with existing CF models consistently outperforms state-of-the-art unbiased recommender models, up to 12.22% for Recall@20 and 16.33% for NDCG@20 gains, on four benchmark datasets.
引用
收藏
页码:2456 / 2460
页数:5
相关论文
共 5 条
  • [1] Towards Representation Alignment and Uniformity in Collaborative Filtering
    Wang, Chenyang
    Yu, Yuanqing
    Ma, Weizhi
    Zhang, Min
    Chen, Chong
    Liu, Yiqun
    Ma, Shaoping
    PROCEEDINGS OF THE 28TH ACM SIGKDD CONFERENCE ON KNOWLEDGE DISCOVERY AND DATA MINING, KDD 2022, 2022, : 1816 - 1825
  • [2] Towards similar alignment and unique uniformity in collaborative filtering
    Sang, Lei
    Zhang, Yu
    Zhang, Yi
    Li, Honghao
    Zhang, Yiwen
    EXPERT SYSTEMS WITH APPLICATIONS, 2025, 259
  • [3] Debiased Contrastive Learning For Graph Collaborative Filtering
    Zhou, Zhijun
    Xie, Qing
    Wang, Yuhan
    Li, Lin
    Liu, Yongjian
    Tang, Mengzi
    PROCEEDINGS OF THE 2024 27 TH INTERNATIONAL CONFERENCE ON COMPUTER SUPPORTED COOPERATIVE WORK IN DESIGN, CSCWD 2024, 2024, : 48 - 54
  • [4] Simplified Graph Contrastive Learning for Recommendation with Direct Optimization of Alignment and Uniformity
    Tian, Renjie
    Jing, Mingli
    Jiao, Long
    Wang, Fei
    ARABIAN JOURNAL FOR SCIENCE AND ENGINEERING, 2024,
  • [5] Category-guided multi-interest collaborative metric learning with representation uniformity constraints
    Wang, Long
    Lian, Tao
    INFORMATION PROCESSING & MANAGEMENT, 2025, 62 (01)