Abstract

Among the flourishing research of weakly supervised learning (WSL), we recognize the lack of a unified interpretation of the mechanism behind the weakly supervised scenarios, let alone a systematic treatment of the risk rewrite problem, a crucial step in the empirical risk minimization approach. In this talk, we introduce a framework providing a comprehensive understanding and a unified methodology for WSL. The formulation component of the framework, leveraging a contamination perspective, provides a unified interpretation of how weak supervision is formed and subsumes fifteen existing WSL settings. The induced reduction graphs offer comprehensive connections over WSLs. The analysis component of the framework, viewed as a decontamination process, provides a systematic method of conducting risk rewrite. In addition to the conventional inverse matrix approach, we devise a novel strategy called marginal chain aiming to decontaminate distributions. We justify the feasibility of the proposed framework by recovering existing rewrites reported in the literature.https://openreview.net/forum?id=RGsdAwWuu6

Bio.

Dr. Chao-Kai Chiang is a Postdoctoral Researcher at National Taiwan University’s AI-CoRE Center, working with Professor Hsuan-Tien Lin in the CLLAB. His research focuses on weakly supervised learning, learning with label noise, and multi-armed bandits, aiming to build reliable and theoretically grounded machine learning methods for real-world data imperfections.

All faculty and students are welcome to join.