Bonus Theory Seminar this Friday: Alex Slivkins

9 views
Skip to first unread message

David Kempe

unread,
Mar 3, 2026, 7:32:13 PM (2 days ago) Mar 3
to usc-theo...@googlegroups.com, USC-T...@googlegroups.com
Hi everyone,

in addition to our regular Thursday noon theory seminar, this week, we
will have a bonus theory seminar on Friday, 03/06/2026, from
2:00-3:00pm. The talk will also be in our new home, GCS 302C.
The speaker will be Alex Slivkins from Microsoft Research.


Title: Bandit Social Learning: Exploration (Failures) under
Exploitation

Abstract: We consider social learning dynamics where the agents
collectively face a multi-armed bandit problem. Self-interested
users/customers ("agents") choose among available alternatives ("arms")
under uncertainty on their quality. The agents return feedback on their
experiences, which is aggregated and served back to the future agents. As
individual agents are reluctant to explore for the sake of others, they may
catastrophically fail to explore as a collective. The same concern underpins
and motivates a huge, decades-long literature on algorithmic exploration in
multi-armed bandits. Yet, this learning dynamics has been surprisingly
poorly understood until recently. We study how it plays out, depending on
the particularities of the agents' behavior and the underlying learning
problem.

Based on two papers: "Bandit Social Learning under Myopic Behavior"
(NeurIPS'23; https://arxiv.org/abs/2302.07425) and  "Greedy Algorithm for
Structured Bandits" (NeurIPS'25; https://arxiv.org/abs/2503.04010).

Speaker: Alex Slivkins
(https://www.microsoft.com/en-us/research/people/slivkins/).

Bio: Alex Slivkins is a Senior Principal Researcher at Microsoft
Research NYC. Some time ago he was a researcher at MSR Silicon
Valley (now defunct), after receiving his Ph.D. from Cornell. His research
interests span learning theory, algorithmic economics, and networks. He is
particularly interested in exploration-exploitation tradeoff and its
manifestations in socioeconomic environments. His work has been recognized
with the best paper award at ACM EC 2010, the best paper nomination at WWW
2015, and the best student paper award at ACM PODC 2005. His book,
"Introduction to Multi-Armed Bandits", has been published (with online
revisions) in 2019-2024.

--
David Kempe <david....@gmail.com>
Reply all
Reply to author
Forward
0 new messages