通过工程活动增强基于 ML 的关键系统可信性
软件工程
2022-10-03 v1 机器学习
摘要
本文综述了为关键系统配备先进分析与决策功能的可信机器学习 (ML) 算法的完整工程过程。我们从 ML 基本原理出发，描述制约其信任的核心要素，尤其通过其设计：即领域规范、数据工程、ML 算法设计、实现、评估与部署。后者被组织为一个用于可信 ML 系统设计的统一框架。
引用
@article{arxiv.2209.15438,
title = {Empowering the trustworthiness of ML-based critical systems through engineering activities},
author = {Juliette Mattioli and Agnes Delaborde and Souhaiel Khalfaoui and Freddy Lecue and Henri Sohier and Frederic Jurie},
journal= {arXiv preprint arXiv:2209.15438},
year = {2022}
}
备注
This work has been supported by the French government under the "France 2030" program, as part of the SystemX Technological Research Institute Research Institute