Hybrid seminar at McGill University or Zoom.

This paper studies learning-based optimal admission control policy for an Erlang-B queuing system under partially observation. At each arrival, the dispatcher observes current occupancy and decides whether to accept or reject the job. A completed job yields a fixed reward but incurs a cost proportional to service duration. The objective is to design an admission control policy maximizing the long-term average reward. Arrival and service rates are unknown, and unobserved departures make parameter estimation difficult. The reward structure induces an asymmetry in the optimal decision across parameter regimes. We establish an instance-dependent asymptotic lower bound, explicit in the system parameters, demonstrating that the worst-case regret is logarithmic in the number of arrivals in one regime; this characterization enables comparison across across regimes. We complement the bound with a matching upper bound via explicit policy construction. We generalize the lower bound to the setting with general service times and using the queueing dynamics devise an order-optimal universal achievable scheme using the one-armed bandit paradigm. We end by discussing how the new achievable scheme extends to more general problems using a connection to the multi-armed bandit problem.

This talk combines joint work with Saghar Adler (TikTok, USDS) and Mehrdad Moharrami (CS, Univ. Of Iowa), and Shubhhi Singh (EECS, Univ. of Michigan) and Shubhanshu Shekhar (EECS, Univ. of Michigan).

Bio: Vijay Subramanian is a Professor in ECE Division of the EECS Department at the University of Michigan, Ann Arbor; from Fall 2014 to Summer 2024 he was an Associate Professor at the same institution. He received the Ph.D. degree in electrical engineering from the University of Illinois at Urbana-Champaign, Champaign, IL, USA, in 1999. Thereafter, he worked at Motorola Inc., the Hamilton Institute, Maynooth, Ireland, and the EECS Department, Northwestern University, Evanston, IL, USA. He also held an Adjunct Assistant Professor position in EECS at Northwestern University from 2014 to 2018, an Adjunct Research Associate Professor in CSL and ECE at UIUC from 2022-2024, and an FSMP Invited Professor Fellowship for INRIA, Paris in 2024. His current research interests are in stochastic analysis, random graphs, multi-agent systems, and game theory (including mechanism and information design) with applications to social, economic and technological networks.