報告題目：Robust machine learning for responsible AI
報告人：王晉東 主管研究員
單位：微軟亞洲研究院
報告時間：2023年3月12日（周日）下午2點
報告地點：翡翠科教樓A座902室
報告摘要：
While the increasing popularity of artificial intelligence (AI) has enriched our daily lives, its responsibility
remains an imminent topic in today’s AI research. Responsible AI involves different aspects such as
transparency, security, robustness, and ethics. In this talk, we will introduce some of our work from the
perspective of robustness, which makes machine learning models more robust to unexpected scenarios.
Specifically, three scenarios got our attention: out-of-distribution generalization to distribution shift, semi-
supervised learning to lowresource labeling environment, and adversarial robustness to malicious attack.
After the introduction of these work, we will also introduce some preliminary robustness analysis to the
recent ChatGPT and large models. Finally, I will discuss some potential research topics in robustness.
報告人簡介：
Dr. Jindong Wang is currently a Senior Researcher at Microsoft Research Asia. He obtained his Ph.D from
Institute of Computing Technology, Chinese Academy of Sciences in 2019. He visited Qiang Yang’s group at
Hong Kong University of Science and Technology in 2018. His research interest includes robust machine
learning, transfer learning, semisupervised learning, and federated learning. He published over 40 papers
with 5000 citations at leading conferences and journals such as ICLR, NeurIPS, CVPR, IJCAI, UbiComp,
ACMMM, TKDE, TASLP etc. He served as the senior program committee member of IJCAI and AAAI, and
PC members for other conferences like ICML, NeurIPS, ICLR, CVPR etc. He opensourced several projects
to help build a better community, such as transferlearning, torchSSL, USB, personalizedFL, and robustlearn,
which received over 10K stars on Github. He published a textbook called Introduction to Transfer Learning
in 2021 to help starters quickly learn transfer learning.
TOP
