arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2502.16263cs.LGcs.CYmath.STstat.MLstat.TH
收藏

基于PLS的公平表示学习方法

PLS-based approach for fair representation learning

Institute of Data Science and Artificial Intelligence (DATAI), Universidad de Navarra(纳瓦拉大学数据科学与人工智能研究所);Universidad Pública de Navarra(纳瓦拉公立大学)

Elena M. De-Diego, Adrián Perez-Suay, Paula Gordaliza, Jean-Michel Loubes

AI总结:

本研究针对公平表示学习问题,提出融入公平性约束的公平PLS成分方法,支持线性及核嵌入下的非线性场景,经多数据集验证,效果优于标准公平PCA方法。

摘要:

我们重新探讨公平表示学习问题,提出了公平偏最小二乘(Fair Partial Least Squares,PLS)成分方法。PLS是统计学中广泛使用的方法,通过生成适配预测任务的表示来高效降维。我们提出一种在PLS成分构建中融入公平性约束的新方法,该算法可在线性与非线性场景下(借助核嵌入)构建此类特征。我们在多个数据集上评估了方法的效率,并证明其性能优于标准公平PCA方法。

英文摘要:

We revisit the problem of fair representation learning by proposing Fair Partial Least Squares (PLS) components. PLS is widely used in statistics to efficiently reduce the dimension of the data by providing representation tailored for the prediction. We propose a novel method to incorporate fairness constraints in the construction of PLS components. This new algorithm provides a feasible way to construct such features both in the linear and the non linear case using kernel embeddings. The efficiency of our method is evaluated on different datasets, and we prove its superiority with respect to standard fair PCA method.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑