Tiny, Always-on, and Fragile: Bias Propagation through Design Choices in On-device Machine Learning Workflows
Tiny, Always-on, and Fragile: Bias Propagation through Design Choices in On-device Machine Learning Workflows
复制标题
微小、始终在线且脆弱：通过设备上机器学习工作流程中的设计选择进行偏差传播
DOI：
10.1145/3591867
复制
发表时间：
2022
影响因子：
4.4
通讯作者：
Akhil Mathur
中科院分区：
文献类型：
--
作者：
W. Hutiri;A. Ding;F. Kawsar;Akhil Mathur
数十亿的分布式、异构和资源受限的物联网设备部署了设备上的机器学习（ML），用于对个人数据进行私有、快速和离线的推断。设备上的ML高度依赖于上下文，并且对用户、使用、硬件和环境属性非常敏感。机器学习中的这种敏感性和偏倚倾向使得研究设备上设置的偏倚变得非常重要。我们的研究是对这一新兴领域偏见的首批调查之一，为构建更公平的设备上机器学习奠定了重要基础。我们应用软件工程的视角，通过设备上机器学习工作流程中的设计选择来研究偏见的传播。我们首先确定了可靠性偏差作为不公平的来源，并提出了一种量化的措施。然后，我们对关键字发现任务进行实证实验，以显示复杂和相互作用的技术设计选择如何放大和传播可靠性偏差。我们的研究结果证实，在模型训练过程中做出的设计选择，如样本率和输入特征类型，以及优化模型的选择，如轻量级架构、修剪学习率和修剪稀疏性，会导致男性和女性群体的预测性能不同。根据我们的研究结果，我们建议工程师采取低努力策略来减轻设备上机器学习的偏见。
Billions of distributed, heterogeneous, and resource constrained IoT devices deploy on-device machine learning (ML) for private, fast, and offline inference on personal data. On-device ML is highly context dependent and sensitive to user, usage, hardware, and environment attributes. This sensitivity and the propensity toward bias in ML makes it important to study bias in on-device settings. Our study is one of the first investigations of bias in this emerging domain and lays important foundations for building fairer on-device ML. We apply a software engineering lens, investigating the propagation of bias through design choices in on-device ML workflows. We first identify reliability bias as a source of unfairness and propose a measure to quantify it. We then conduct empirical experiments for a keyword spotting task to show how complex and interacting technical design choices amplify and propagate reliability bias. Our results validate that design choices made during model training, like the sample rate and input feature type, and choices made to optimize models, like light-weight architectures, the pruning learning rate, and pruning sparsity, can result in disparate predictive performance across male and female groups. Based on our findings, we suggest low effort strategies for engineers to mitigate bias in on-device ML.
DOI：
10.1145/3442188.3445865
发表时间：
2019-11
期刊：
Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency
影响因子：
--
作者：
Harvineet Singh;Rina Singh;Vishwali Mhasawade;R. Chunara
通讯作者：
Harvineet Singh;Rina Singh;Vishwali Mhasawade;R. Chunara
DOI：
--
发表时间：
2021-06
期刊：
ArXiv
影响因子：
--
作者：
Colby R. Banbury;V. Reddi;P. Torelli;J. Holleman;Nat Jeffries;C. Király;Pietro Montino;David Kanter;S. Ahmed;Danilo Pau;Urmish Thakker;Antonio Torrini;Pete Warden;Jay Cordaro;G. D. Guglielmo;Javier Mauricio Duarte;Stephen Gibellini;Videet Parekh;Honson Tran;Nhan Tran;Niu Wenxu;Xu Xuesong
通讯作者：
Colby R. Banbury;V. Reddi;P. Torelli;J. Holleman;Nat Jeffries;C. Király;Pietro Montino;David Kanter;S. Ahmed;Danilo Pau;Urmish Thakker;Antonio Torrini;Pete Warden;Jay Cordaro;G. D. Guglielmo;Javier Mauricio Duarte;Stephen Gibellini;Videet Parekh;Honson Tran;Nhan Tran;Niu Wenxu;Xu Xuesong
构图下的公平性
DOI：
--
发表时间：
2019
期刊：
10th Innovations in Theoretical Computer Science Conference (ITCS 2019
影响因子：
--
作者：
Dwork, C;Ilvento, C
通讯作者：
Ilvento, C