Sequence Unlearning for Sequential Recommender Systems

被引:0
|
作者
Ye, Shanshan [1 ]
Lu, Jie [1 ]
机构
[1] Univ Technol Sydney, Australian Artificial Intelligence Inst, Fac Engn & IT, Sydney, NSW, Australia
来源
ADVANCES IN ARTIFICIAL INTELLIGENCE, AI 2023, PT I | 2024年 / 14471卷
关键词
Sequential Recommender Systems; Machine Unlearning; Noise Injection;
D O I
10.1007/978-981-99-8388-9_33
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Sequential recommender systems, leveraging clients' sequential product browsing history, have become an essential tool in delivering personalized product recommendations. As data protection regulations come into focus, certain clients may demand the removal of their data from the training sets used by these systems. In this paper, we focus on the problem of how specific client information can be efficiently removed from a pre-trained sequential recommender system without the need for retraining, particularly when the change to the data set is not substantial. We propose a novel sequence unlearning method for sequential recommender systems by leveraging label noise injection. Intuitively, our method promotes data unlearning by encouraging the system to produce random predictions for the sequences aiming to unlearn. To further prevent the model from overfitting an incorrect label, which could lead to substantial changes in its parameters, our method incorporates a dynamic process wherein the incorrect label is continually altered during the learning phase. This effectively encourages the model to lose confidence in the original label, while also discouraging it from fitting to a specific incorrect label. To the best of our knowledge, this is the first work to tackle the unlearning problem in sequential recommender systems without accessing the remaining data. Our approach is general and can work with any sequential recommender system. Empirically, we demonstrate that our method effectively helps different recommender systems unlearn specific sequential data while maintaining strong generalization performance on the remaining data across different datasets.
引用
收藏
页码:403 / 415
页数:13
相关论文
共 50 条
  • [31] Improving Conversational Recommender Systems via Transformer-based Sequential Modelling
    Zou, Jie
    Kanoulas, Evangelos
    Ren, Pengjie
    Ren, Zhaochun
    Sun, Aixin
    Long, Cheng
    PROCEEDINGS OF THE 45TH INTERNATIONAL ACM SIGIR CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION RETRIEVAL (SIGIR '22), 2022, : 2319 - 2324
  • [32] LinRec: Linear Attention Mechanism for Long-term Sequential Recommender Systems
    Liu, Langming
    Cai, Liu
    Zhang, Chi
    Zhao, Xiangyu
    Gao, Jingtong
    Wang, Wanyu
    Lv, Yifu
    Fan, Wenqi
    Wang, Yiqi
    He, Ming
    Liu, Zitao
    Li, Qing
    PROCEEDINGS OF THE 46TH INTERNATIONAL ACM SIGIR CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION RETRIEVAL, SIGIR 2023, 2023, : 289 - 299
  • [33] A Survey and Taxonomy of Sequential Recommender Systems for E-commerce Product Recommendation
    Nasir M.
    Ezeife C.I.
    SN Computer Science, 4 (6)
  • [34] The case for unlearning health systems
    Joo, Joseph H.
    Kim, Joshua A.
    Liao, Joshua M.
    HEALTHCARE-THE JOURNAL OF DELIVERY SCIENCE AND INNOVATION, 2022, 10 (04):
  • [35] DSRS: DELIGHT sequential recommender system
    Ullah, Syed Tauhid
    Khan, Fazlullah
    Yamani, Shirin
    Alturki, Ryan
    Gazzawe, Foziah
    Razzak, Muhammad Imran
    ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE, 2025, 142
  • [36] T3SRS: Tensor Train Transformer for compressing sequential recommender systems
    Li, Hao
    Zhao, Jianli
    Huo, Huan
    Fang, Sheng
    Chen, Jianjian
    Yao, Lutong
    Hua, Yiran
    EXPERT SYSTEMS WITH APPLICATIONS, 2024, 238
  • [37] Sequential Learning over Implicit Feedback for Robust Large-Scale Recommender Systems
    Burashnikova, Aleksandra
    Maximov, Yury
    Amini, Massih-Reza
    MACHINE LEARNING AND KNOWLEDGE DISCOVERY IN DATABASES, ECML PKDD 2019, PT III, 2020, 11908 : 253 - 269
  • [38] TruthSR: Trustworthy Sequential Recommender Systems via User-generated Multimodal Content
    Yan, Meng
    Huang, Haibin
    Liu, Ying
    Zhao, Juan
    Gao, Xiyue
    Xu, Cai
    Guan, Ziyu
    Zhao, Wei
    DATABASE SYSTEMS FOR ADVANCED APPLICATIONS, DASFAA 2024, PT 3, 2025, 14852 : 180 - 195
  • [39] Recommender systems
    Resnick, P
    Varian, HR
    COMMUNICATIONS OF THE ACM, 1997, 40 (03) : 56 - 58
  • [40] Recommender systems
    Felfernig, Alexander
    Friedrich, Gerhard
    Schmidt-Thieme, Lars
    IEEE INTELLIGENT SYSTEMS, 2007, 22 (03) : 18 - 21