报告时间：2023年12月5日（星期二）上午11:00

报告地点：华南理工大学南校区软件学院B7-403

报告题目：Deep Disruption-robust Machine Intelligence in Open Environments

报告人：Dr. Yanjie Fu

主持人：杨晓伟教授

报告摘要：

AI systems have been increasingly deployed into real world infrastructures (e.g., power grid load forecasting and management, cancer detection, autonomous vehicles). Can they be safe and trusted to perform what they are designed to do even under real world disruptions (e.g., volatility, bias, uncertainty, distribution shift, disorder, mistakes, faults, attacks, or failures)?  AI systems, unlike humans, are brittle, fragile, often struggle when faced with novel, dynamic, and uncertain situations (a.k.a., open environments), and highly sensitive to small perturbations, which can lead to catastrophically poor performance.

In this talk, I will discuss our thinking on disruption-robust AI in open environments. Under the disruption robustness context, we see AI systems as a joint objective of maximizing task success and minimizing negative environment impacts. Our perspective is to connect methodological and computing issues in major machine learning paradigms, e.g., representation learning (imperfect data, structure knowledge), self-supervised learning (limitation of labels), interactive learning (weak supervision and constrained environments), adaptive learning (uncertain and drifting environment), stream learning (continuous learning and limitation of memory), as disruption-robust learning. I will discuss: 1) how to construct robust data representation to fight data disruptions (imperfect data and complex knowledge structure); 2) how to construct robust learning strategies to fight machine learning environment disruptions (uncertain and constrained environments). Finally, I will include the key research insights and present future work.

个人简介：

Dr. Yanjie Fu is an associate professor in the School of Computing and AI at the Arizona State University. He received his Ph.D. degree from the Rutgers, the State University of New Jersey in 2016, the B.E. degree from the University of Science and Technology of China in 2008, and the M.E. degree from the Chinese Academy of Sciences in 2011. He has research experience in industry research labs, such as Microsoft Research Asia and IBM Thomas J. Watson Research Center. He has published prolifically in refereed journals and conference proceedings, such as IEEE TKDE, IEEE TMC, ACM TKDD, ACM SIGKDD, AAAI, IJCAI, VLDB, WWW, ACM SIGIR. His research has been recognized by: 1) two federal junior faculty awards: US NSF CAREER and NSF CRII awards; 2) five best paper (runner-up, finalist) awards, including ACM KDD18 Best Student Paper Finalist, IEEE ICDM14, 21, 22 Best Paper Finalist, ACM SIGSpatial20 Best Paper Runner-up; 3) three industrial awards: 2016 Microsoft Azure Research Award, 2022 Baidu Scholar global top Chinese young scholars in AI, 2021 Aminer.org AI 2000 Most Influential Scholar Award Honorable Mention in Data Mining; 4) several other university-level awards: Reach the Stars Award, University System Research Board Award and University Interdisciplinary Research Award. He was chosen for the nation’s early career engineers by the US National Academy of Engineering 2023 Grainger Foundation Frontiers of Engineering Symposium. He is committed to data science education. His graduated Ph.D. students have joined academia as tenure-track faculty members. He is broadly interested in data mining, machine learning, and their interdisciplinary applications. His research aims to develop robust machine intelligence with imperfect and complex data by building tools to address framework, algorithmic, data, and computing challenges. His recent focuses are spatial-temporal AI, graph learning, reinforcement learning, learning with unlabeled data, stream learning and distribution drift. He currently serves as an Associate Editor of ACM Transactions on Knowledge Discovery from Data and Mathematics. He is a senior member of ACM and IEEE.

总编：黄翰

责任编辑：袁中锦

时间：2023年12月3日