Decoupled Classifiers for Group-Fair and Efficient Machine Learning
Decoupled Classifiers for Group-Fair and Efficient Machine Learning
复制标题
用于群体公平和高效机器学习的解耦分类器
DOI：
--
复制
发表时间：
2017
期刊：
影响因子：
--
通讯作者：
Mark D. M. Leiserson
中科院分区：
文献类型：
--
作者：
C. Dwork;Nicole Immorlica;A. Kalai;Mark D. M. Leiserson
当在机器学习系统中使用敏感属性(如性别或种族)是合乎道德和合法的时候，问题仍然是如何做到这一点。我们表明，使用敏感属性的机器学习算法的朴素应用导致了组之间固有的精度交易ff。我们提供了一种简单而有效的解耦技术，它可以添加到任何黑盒机器学习算法之上，来学习不同类别的不同类别(Diffentffifffier)。转移学习被用来缓解任何一个组的数据太少的问题。
When it is ethical and legal to use a sensitive attribute (such as gender or race) in machine learning systems, the question remains how to do so. We show that the na¨ıve application of machine learning algorithms using sensitive attributes leads to an inherent tradeoff in accuracy between groups. We provide a simple and efficient decoupling technique, which can be added on top of any black-box machine learning algorithm, to learn different classifiers for different groups. Transfer learning is used to mitigate the problem of having too little data on any one group.