Consider an RMAB instance with $N$ arms, where each arm $i \in [N]$ has a finite state space $\mathbb S_i$ and can receive an action $y_i^t \in {0, 1}$ (representing not pulling or pulling the arm, respectively) at each time step $t$. The state of arm $i$ at time $t$ is denoted by $s_i^t$. Depending on the action taken, a reward $r_i(s_i^t, y_i^t)$ is accrued. As a decision maker, our objective is to maximize the averaged total reward over an infinite time horizon, under a constraint that only $B$ arms can be pulled at any time step....

Decision-making in dynamic environments over time presents unique challenges, particularly when the conditions influencing decisions are constantly changing. In such scenarios, traditional decision-making models often fall short. This is where Restless Multi-Armed Bandits (RMAB) come into play. The method provides a robust framework for modeling and optimizing decisions over time. Here’s a brief introduction of the concept of RMAB and the Whittle Index Policy. understanding Restless Multi-Armed Bandits what are RMAB?...

A memoir of Market Mechanism Design course’s final presentation report, based on: Algorithmic Classification and Strategic Effort Jon Kleinberg and Manish Raghavan | ACM SIGecom Exchanges, Vol. 18, No. 2, November 2020, Pages 53–60 motivation: difference in modelling strategic behavior and objectives–between econ/CS perspectives The principal-agent and strategic machine learning literatures appear to share a common goal: how should one structure a decision-making rule to account for the strategic actions of decision subjects?...

speaker: Zizhuo Wang | Prof., The Chinese University of Hong Kong-Shenzhen. Takeaway: point of view of bridging practice and research. For industry solutions: Need to consider a lot of details Need to be fast and intepretation, therefore requires simple solutions Don’t care about theory. For academic works: Need abstract to focus Graced with time Expect some generality, therefore theory. TALK ABSTRACT: In the past years, there have been growing number of companies in China that adopt OR methods in their operations....

speaker: Ming Hu | Prof., University of Toronto Keys: random product offering/demand allocation/pricing algorithm. TALK ABSTRACT: First, we study blind boxes as a novel selling mechanism in which buyers purchase sealed packages containing unknown items, with the chance of uncovering rare or special items. We show how such product randomization introduced by the blind box can improve the seller’s profitability over traditional separate selling. Second, we study how an e-commerce platform should assign sequentially arriving customers to sellers who compete to sell identical products on the platform....

speaker: Peng Sun | Prof., Duke University TALK ABSTRACT: New prescription drugs require regulatory approval before drug makers can sell them. In some countries, regulators may conditionally approve a drug, which allows sales to begin before the developer has proven the drug’s efficacy. After further testing, the regulator may either grant final approval or reject the drug. We show that conditional approval not only speeds access to drugs but also encourages the development of drugs that would not have been pursued otherwise....

speaker: Ningyuan Chen | Prof., University of Toronto related paper: Allocating Divisible Resources on Arms with Unknown and Random Rewards TALK ABSTRACT: We consider a decision maker allocating one unit of renewable and divisible resource in each period on a number of arms. The arms have unknown and random rewards whose means are proportional to the allocated resource and whose variances are proportional to an order b of the allocated resource....

speaker: Guillermo Gallego | Prof., The Chinese University of Hong Kong-Shenzhen. TALK ABSTRACT: To study the limits of personalization, we introduce the notion of a clairvoyant firm that can read the mind of consumers and sell them the highest revenue product that they are willing to buy. We show how to compute the expected revenue of the clairvoyant firm for a class of rational discrete choice models, and develop prophet-type inequalities that provide performance guarantees for the expected revenue of the traditional assortment optimization firm (a TAOP firm) relative to the clairvoyant firm, and therefore to any effort to personalize assortments....

What an elegant name! It reminded me of the book Mostly Harmless Econometrics. Mostly OM is an annual international workshop sponsored by the School of Economics and Management, the Research Center for Contemporary Management, of Tsinghua University, focusing on the state-of-the-art research in Operations Management (OM), broadly defined. Mostly OM aims to provide a forum for researchers, from China and overseas alike, to foster interaction and exchange ideas, learn from each other new methodologies and applications, and explore opportunities for collaboration....

Andante cantabile(It.) is a direction often used by composers. In general, it means flowing and songlike. The andante tempo is known as the walking pace, and the pacing and energy should generally feel like a nice stroll. Cantabile [kanˈtaːbile], an Italian word, means literally “singable” or “songlike”. In instrumental music, it is a particular style of playing designed to imitate the human voice. To a large section of the public, however, it means one work, the 2nd movt....