Machine Learning from Imperfect Supervision

Masashi Sugiyama

Presentation information

[J] Oral

U (Union ) » Union

[U-13] Future of Earth and Planetary Sciences Boosted by Artificial Intelligence

Sun. May 25, 2025 10:45 AM - 12:15 PM Exhibition Hall Special Setting (1) (Exhibition Hall 7&8, Makuhari Messe)

convener:Hiromichi Nagao(Earthquake Research Institute, The University of Tokyo), Yukihiro Takahashi(Department of Cosmosciences, Graduate School of Science, Hokkaido University), Yusuke Iida(Niigata University), Masuo Nakano(Japan Agency for Marine-Earth Science and Technology), Chairperson:Hiromichi Nagao(Earthquake Research Institute, The University of Tokyo), Yukihiro Takahashi(Department of Cosmosciences, Graduate School of Science, Hokkaido University), Yusuke Iida(Niigata University), Masaru Nakano(Japan Agency for Marine-Earth Science and Technology)

10:45 AM - 11:15 AM

[U13-01] Machine Learning from Imperfect Supervision

★Invited Papers

*Masashi Sugiyama¹ (1.RIKEN/The University of Tokyo)

Keywords:machine learning, weakly supervised learning, label-noise robust learning, transfer learning

In recent applications of machine learning, it is often challenging to collect high-quality labeled data. This talk will introduce weakly supervised learning methods, label-noise robust learning techniques, and transfer learning approaches that enable accurate learning even from incomplete labeled data.