Lecture Introduction:

This short course aims to introduce graduate students to the fundamental concepts, algorithms, and analytical tools of online learning, and to demonstrate how these methods can be applied to solve dynamic decision-making problems in economics. Through this course, students will:

• 1.Understand the core framework of sequential decision making under uncertainty, including the notions of regret, feedback models, and adversarial versus stochastic environments.
• 2.Learn key algorithms in online learning, such as Follow-the-Leader, Follow-the-Perturbed-Leader, Hedge, and Upper Confidence Bound (UCB), and understand their performance guarantees.
• 3.Develop an appreciation for the exploration–exploitation trade-off and how structural assumptions (such as Lipschitz continuity or contextual information) affect algorithmic design.
• 4.Recognize the connections between learning and economic models, including applications to dynamic pricing, online auctions, contract design, and information design.
• 5.Acquire foundational skills to read and analyze current research papers in learning theory, algorithmic economics, and online optimization.

By the end of the course, students are expected to be able to formalize online learning problems, understand regret analyses at a conceptual level, and identify how learning techniques can be integrated into economic and market-design problems.

授课教师:

冯逸丁，香港科技大学（HKUST）工业工程与决策分析助理教授。当前的研究侧重于在线市场中的算法和策略开发。主要利用在线算法、近似算法、机制设计和信息设计的方法来解决相关问题。

日程:

第14周，星期一第5-8节 星期三第5-8节 星期四第5-8节 星期五第5-8节 (立人楼A101)