Date and Time: October 13th, 2026, 9:15 -- 10:15 (JST)
Venue: Hybrid: Online and Open Space at the RIKEN Nihonbashi Office
Title: Connecting Domain Adaptation to Other Learning Problems
Speaker: Prof. Hsuan-Tien Lin (National Taiwan University)
Abstract:
Domain adaptation (DA) has accumulated a large toolbox of domain-specific alignment techniques, but some of the most useful progress comes from recognizing that DA is, in disguise, a problem the machine learning community already knows how to solve. In this talk, I illustrate this perspective through two recent works from my group. The first connects semi-supervised DA to learning with noisy labels. When source and target semantics disagree, aligning target data to source data drags examples toward the wrong classes, so we instead treat source labels as a noisy version of the ideal target labels. The resulting Source Label Adaptation framework (CVPR 2023) cleans source labels from the target's point of view, and because it attacks an orthogonal facet of the problem, it composes with existing DA algorithms to improve them.
The second connects universal DA to representation learning. In the under-explored "extreme" regime where many source classes are absent from the target, the dominant partial-domain-matching paradigm surprisingly falls below a source-only baseline. We trace this failure to dimensional collapse of target representations, and show that alignment and uniformity objectives from self-supervised learning, tools built for representation quality rather than domain transfer, restore the structure that domain matching depends on, yielding state-of-the-art results on a broader benchmark (ICML 2025). Together, these results suggest a general strategy: when DA is hard, connecting it to established solutions in other learning problems.
Short Bio: Prof. Hsuan-Tien Lin received his B.S. in Computer Science and Information Engineering from National Taiwan University in 2001, and his M.S. and Ph.D. in Computer Science from California Institute of Technology in 2005 and 2008, respectively. He joined the Department of Computer Science and Information Engineering at National Taiwan University as an Assistant Professor in 2008, was promoted to Associate Professor in 2012, and has been a Professor since August 2017. In 2022, he was named the Cyberlink/Perfect Endowed Chair Professor; in 2025, he was named the Quanta Endowed Chair Professor. From 2016 to 2019, he served as Chief Data Scientist at Appier, a startup company that specializes in making AI easier to use across domains such as digital marketing and business intelligence, and he continued as its Chief Data Science Consultant until 2025.
From the university, Prof. Lin received the Distinguished Teaching Awards in 2011, 2021, and 2026, earning him the recognition of Lifetime Distinguished Teaching. He received another lifetime recognition, the Outstanding Mentoring Award, in 2013. He also received five Outstanding Teaching Awards between 2016 and 2020. He co-authored the introductory machine learning textbook Learning from Data and offered two popular Mandarin-taught MOOCs, Machine Learning Foundations and Machine Learning Techniques, based on the textbook.
Prof. Lin served in the machine learning community as Program Co-Chair of NeurIPS 2020, Expo Co-Chair of ICML 2021, Workshop Co-Chair of NeurIPS 2022, Workshop Chair of NeurIPS 2023, Program Co-Chair of ACML 2024, General Co-Chair of ACML 2025, Senior Program Chair of NeurIPS 2025, and General Co-Chair of NeurIPS 2026. His research interests include mathematical foundations of machine learning, studies on new learning problems, and improvements to learning algorithms. He received the 2012 K.-T. Li Young Researcher Award from the ACM Taipei Chapter, the 2013 D.-Y. Wu Memorial Award from the National Science Council of Taiwan, the 2017 Creative Young Scholar Award from the Foundation for the Advancement of Outstanding Scholarship in Taiwan, the 2025 Breakthrough Chair Professorship from the Foundation for the Advancement of Outstanding Scholarship in Taiwan, and the 2025 Outstanding Research Award from the National Science and Technology Council of Taiwan. He co-led the teams that won third place in the slow track of KDDCup 2009, the championship of KDDCup 2010, the double championship in both tracks of KDDCup 2011, the championship of track 2 in KDDCup 2012, and the double championship in both tracks of KDDCup 2013.
