Skip to main content
QUICK REVIEW

[论文解读] Improving Recommendation Fairness via Data Augmentation

Lei Chen, Wu Le|arXiv (Cornell University)|Feb 13, 2023
Recommender Systems and Techniques被引用 6
一句话总结

该论文提出FDA,一种数据增强框架,通过基于两个假设生成合成用户-物品交互来提升推荐公平性:一个群体中的用户与另一群体中的用户具有相似的物品偏好。通过使用两阶段优化生成更公平的训练数据,FDA在不依赖预定义公平性度量的情况下提升了公平性,且在真实世界数据集上的多个推荐模型中实现了更优的准确率-公平性权衡。

ABSTRACT

Collaborative filtering based recommendation learns users' preferences from all users' historical behavior data, and has been popular to facilitate decision making. R Recently, the fairness issue of recommendation has become more and more essential. A recommender system is considered unfair when it does not perform equally well for different user groups according to users' sensitive attributes~(e.g., gender, race). Plenty of methods have been proposed to alleviate unfairness by optimizing a predefined fairness goal or changing the distribution of unbalanced training data. However, they either suffered from the specific fairness optimization metrics or relied on redesigning the current recommendation architecture. In this paper, we study how to improve recommendation fairness from the data augmentation perspective. The recommendation model amplifies the inherent unfairness of imbalanced training data. We augment imbalanced training data towards balanced data distribution to improve fairness. The proposed framework is generally applicable to any embedding-based recommendation, and does not need to pre-define a fairness metric. Extensive experiments on two real-world datasets clearly demonstrate the superiority of our proposed framework. We publish the source code at https://github.com/newlei/FDA.

研究动机与目标

  • 解决由训练数据不平衡导致的协同过滤推荐系统中的公平性差异问题。
  • 在不依赖预定义公平性度量或架构修改的情况下提升公平性。
  • 探索数据增强是否能有效缓解推荐系统中的基于群体的公平性偏差。
  • 开发一种适用于任何基于嵌入的推荐模型的通用框架。

提出的方法

  • 提出两个假设:(1) 一个群体中的用户与另一群体中的用户具有相似的点击/非点击偏好;(2) 非点击偏好也适用相同假设。
  • 利用这些假设生成合成(虚假)用户-物品交互,以平衡敏感群体之间的训练数据分布。
  • 采用两阶段优化:内层循环生成改进的虚假数据,外层循环在增强数据(真实 + 虚假)上训练推荐模型。
  • 将该框架应用于任何基于嵌入的推荐模型(如BPR和GCCF),无需修改模型架构。
  • 使用Jensen-Shannon散度衡量公平性,通过比较用户群体在Top-K推荐中物品分布的差异。
  • 无需显式公平性正则化或敏感属性解耦,使其与模型无关且易于部署。
Figure 1. The overall architecture of the proposed FDA framework. By not changing the recommendation model, our key idea is to improve data balance between different user groups via data augmentation. For each real data $(u_{0},i,r)$ of user $u_{0}$ ( $u_{0}\in G_{0}$ ), item $i$ , and the implicit
Figure 1. The overall architecture of the proposed FDA framework. By not changing the recommendation model, our key idea is to improve data balance between different user groups via data augmentation. For each real data $(u_{0},i,r)$ of user $u_{0}$ ( $u_{0}\in G_{0}$ ), item $i$ , and the implicit

实验结果

研究问题

  • RQ1在不修改推荐模型的情况下,数据增强是否能有效减少推荐系统中的公平性差异?
  • RQ2所提出的跨群体偏好相似性假设对公平性提升有何影响?
  • RQ3该两阶段优化框架在多大程度上能平衡不同模型的推荐准确率与公平性?
  • RQ4生成的虚假交互数量在多大程度上影响公平性-准确率权衡?

主要发现

  • FDA显著降低了MovieLens和LastFM数据集上由Jensen-Shannon散度衡量的公平性差异。
  • 在LastFM数据集上,FDA将Top-20推荐中JS散度从BPR的0.4842降低至0.3618,从GCCF的0.4879降低至0.3721,显示出显著的公平性提升。
  • 在LastFM上,BPR的最优虚假数据占比为0.3,GCCF为0.4,该比例在公平性与准确率之间实现了良好平衡。
  • 消融实验证实,两个假设独立且共同发挥作用,完整框架优于单一假设的变体。
  • FDA在显著提升公平性的同时,保持或略微提升推荐准确率,即使基础模型(BPR)本身准确率较低。
  • 该框架在不同推荐主干模型(BPR、GCCF)上均表现良好,证实了其通用性与鲁棒性。
Figure 2. Performance under different ratios on LastFM.
Figure 2. Performance under different ratios on LastFM.

更好的研究,从现在开始

从阅读论文到最终审阅,大幅缩短您的研究时间。

无需绑定信用卡

本解读由 AI 生成,并经人工编辑审核。