Interpretability for Machine Learning

机器学习的可解释性

• 批准号: CRC-2019-00383
• 负责人: Matwin, Stan
• 金额: $ 3.64万
• 依托单位: Dalhousie University
• 依托单位国家: 加拿大
• 项目类别: Canada Research Chairs
• 财政年份: 2019
• 资助国家: 加拿大
• 起止时间: 2019-01-01 至 2020-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=689994
• 关键词: Interpretability; Machine; Learning

Recent important breakthroughs of modern Machine Learning (ML) and Deep Learning methods are accompanied by the concerns about transparency and fairness of ML systems. In this context interpretability, understood as the ability to explain and present ML systems to humans, becomes an important requirement for modern ML systems. This research program will tackle several important aspects of interpretability, including novel use of relational knowledge representation techniques in which I have significant accomplishments. The programs principal axes are the involvement of graduate students, as well as collaborative applications of Machine Learning in Dalhousie Universitys priority research areas of Health and Oceans.

现代机器学习（ML）和深度学习方法最近取得的重要突破伴随着对ML系统透明度和公平性的担忧。在这种情况下，可解释性，理解为向人类解释和呈现ML系统的能力，成为现代ML系统的重要要求。这个研究项目将解决可解释性的几个重要方面，包括关系知识表示技术的新用途，我在这方面取得了重大成就。该计划的主轴是研究生的参与，以及机器学习在达尔豪西大学的健康和海洋的优先研究领域的合作应用。