Rényi Fair Inference
Rényi Fair Inference
复制标题
Rényi 公平推理
DOI：
--
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
Meisam Razaviyayn
中科院分区：
文献类型：
--
作者：
Sina Baharlouei;Maher Nouiehed;Meisam Razaviyayn
机器学习算法已越来越多地部署在直接影响人类生活的关键自动化决策系统中。当这些算法仅经过训练以最小化训练/测试误差时，它们可能会因性别或种族等敏感属性而遭受针对个人的系统性歧视。最近，机器学习界掀起了一股开发公平机器学习算法的热潮。特别是，已经提出了许多对抗性学习程序来强加公平性。不幸的是，这些算法要么只能在变量之间强加一阶依赖的公平性，要么缺乏计算收敛保证。在本文中，我们使用 Renyi 相关性作为机器学习模型公平性的衡量标准，并开发了一个通用的训练框架来实现公平性。特别是，我们提出了一种最小-最大公式，在求解最优性时平衡准确性和公平性。对于离散敏感属性的情况，我们提出了一种具有理论收敛保证的迭代算法来解决所提出的最小-最大问题。然后，我们的算法和分析专门用于不同影响原则下的公平分类和公平聚类问题。最后，在 Adult 和 Bank 数据集上评估了所提出的 Renyi 公平推理框架的性能。
Machine learning algorithms have been increasingly deployed in critical automated decision-making systems that directly affect human lives. When these algorithms are only trained to minimize the training/test error, they could suffer from systematic discrimination against individuals based on their sensitive attributes such as gender or race. Recently, there has been a surge in machine learning society to develop algorithms for fair machine learning. In particular, many adversarial learning procedures have been proposed to impose fairness. Unfortunately, these algorithms either can only impose fairness up to first-order dependence between the variables, or they lack computational convergence guarantees. In this paper, we use Renyi correlation as a measure of fairness of machine learning models and develop a general training framework to impose fairness. In particular, we propose a min-max formulation which balances the accuracy and fairness when solved to optimality. For the case of discrete sensitive attributes, we suggest an iterative algorithm with theoretical convergence guarantee for solving the proposed min-max problem. Our algorithm and analysis are then specialized to fair classification and the fair clustering problem under disparate impact doctrine. Finally, the performance of the proposed Renyi fair inference framework is evaluated on Adult and Bank datasets.
DOI：
10.1609/aaai.v34i04.6002
发表时间：
2019-03
期刊：
--
影响因子：
--
作者：
Ashkan Rezaei;Rizal Fathony;Omid Memarrast;Brian D. Ziebart
通讯作者：
Ashkan Rezaei;Rizal Fathony;Omid Memarrast;Brian D. Ziebart
DOI：
--
发表时间：
2019-02
期刊：
ArXiv
影响因子：
--
作者：
A. Backurs;P. Indyk;Krzysztof Onak;B. Schieber;A. Vakilian;Tal Wagner
通讯作者：
A. Backurs;P. Indyk;Krzysztof Onak;B. Schieber;A. Vakilian;Tal Wagner