講者：Wu Yi-Shan博士後研究員（中央研究院資訊科技創新研究中心）
時間：115年9月17日（星期四）10:30
地點：台大次震宇宙館601室
講題：Learning from Data, Feedback, and Interaction: Statistical Principles for Prediction and Sequential Decision-Making
摘要：Learning problems differ in how data are collected and how a learner’s decisions affect future observations. Across batch learning, bandits, and reinforcement learning, my research asks what structure connects the quantities a learner can estimate or optimize to the performance that matters, such as risk, regret, or return. I will present examples from my work showing how different learning protocols change the assumptions and guarantees behind familiar statistical tools, and how these changes challenge algorithm design.