CLIMB talk with Amin Saberi

I will explore the emerging theory of optimal online policies in matching and allocation problems. In many online optimization settings, the standard benchmark is the optimal offline policy—as exemplified by the long-studied prophet inequalities, which have strongly influenced modern work on online markets and mechanism design. Yet an equally natural benchmark in stochastic settings is the optimal online policy, long considered elusive, with the consensus that "we do not have many techniques for getting a handle on it."

That is changing. Recent breakthroughs now allow us to characterize and approximate the optimal online policy, uncovering fundamental structure in online decision-making under uncertainty. This talk will survey these advances—spanning approximation, learning, and mechanism design—and conclude with a discussion of open questions and future directions in this rapidly evolving field.