Date/Time
Date(s) - March 21, 2022
11:45 am - 12:35 pm
Location
406 Weil Hall
Categories
Shuai Huang, Ph.D.
Associate Professor, Department of Industrial & Systems Engineering
University of Washington
Title: From Rule-based Discovery Tool to Collaborative Learning Framework: A Story about How Interpretable and Fairness-aware AI can Help Battle Type 1 Diabetes.
Abstract: The last few years have witnessed many fruitful discussions among public sector organizations, research institutions, and private companies about what are the principles and guidelines for ethic AI practices. It is frequently reported that these principles are not always compatible with each other, so tradeoffs are expected, and which principles weight more is often context-dependent or task-dependent. Nonetheless, one consensus is that AI needs to be interpretable and trustworthy, so it is crucial that people can understand how an AI model works and interpret its decisions. This interpretability of an AI model is not only a practical need but also forms the technical foundation of other ethic dimensions such as fairness, accountability, and privacy. In this talk, I will reinforce this observation and contribute with new insights by sharing my research on Type 1 diabetes (T1D) for which I have been contributing novel AI models and tools such as the rule-based interpretable knowledge discovery tool and the collaborative learning method that can mitigate a range of data disparities. I will give a historic account of this 10-year journey to show how the interpretability of these AI models was gradually developed over time, shaped by a close collaboration between both sides who are habituated to tackle problems with familiar angles and established frameworks but compelled to find new interdisciplinary solutions for challenging problems. I will also talk about my ongoing works to combine the strengths of rule-based method with collaborative learning to create interpretable, transparent, and fairness-enforced risk models of T1D to overcome issues like minority bias, aggregation bias, and sampling bias, and achieve equitable performance across multiple ethnic groups. Finally, I will also talk about how to design a curriculum to better prepare our ISE students for a new AI-enabled workforce whose mission is to install AI into our economy, culture, and society in the coming decades.
About Shuai Huang, Ph.D.
Dr. Shuai Huang is an Associate Professor at the Department of Industrial & Systems Engineering at the University of Washington. He received a B.S. degree on Statistics from the School of Gifted Young at the University of Science and Technology of China in 2007 and a Ph.D. degree on Industrial Engineering from the Arizona State University in 2012. Shuai’s research is driven by challenging data analytics and AI problems, emphasizes innovation in machine learning and AI modeling for complex systems and processes in the connected world, automates the integration of human with these data-driven learning systems, and targets interpretable and explainable decision-makings with discretion of AI ethics and accountability. He develops methodologies for modeling, monitoring, diagnosis, and prognosis of complex networked systems such as brain connectivity networks, cyber physical systems, disease progression processes, and many emerging applications in IoT. He also develops machine learning and AI models to integrate massive and heterogeneous datasets such as neuroimaging, genomics, proteomics, laboratory tests, demographics, and clinical variables, for facilitating scientific discoveries in biomedical research and better decision-makings in clinical practices. His research has been funded by the National Science Foundation (NSF), National Institute of Health (NIH), Defense Advanced Research Projects Agency (DARPA), Air Force Office of Scientific Research (AFOSR), Juvenile Diabetes Research Foundation (JDRF), and several other research institutes and foundations. Dr. Huang currently serves as Associate Editor for the IIE Transactions in Healthcare Systems Engineering and INFORMS Journal of Data Science.
Attend this seminar virtually:
https://ufl.zoom.us/j/99952884272?pwd=aGRCOGZqdk9xcDkzMlVkSHgzY29Pdz09
Meeting ID: 999 5288 4272
Passcode: 452792