Upcoming Seminars

Abstract

We initialize the study of all-pairs shortest distances of a graph in the continual release streaming model under differential privacy for a fixed graph topology with private edge weights. We first give an ε-differentially private mechanism for answering L shortest distance queries per timestep over time horizon T with worst-case error of O(\sqrt{LT/ε}). We then show that this result can be improved for well-structured graphs by introducing the notion of a summary set for the set P of all paths between the L pairs of vertices, which identifies the commonly used subpaths on the graphs, for which privacy noise can be added once and re-used when computing shortest distances on graphs. We prove that good summary sets imply good error upper bounds for the differentially private all-pairs shortest distances problem. In particular, for answering L shortest distance queries per time step, the existence of a (k, d)-summary set S of P implies a ε-differentially private mechanism with worst-case error of O(kd log |S| log^2 T/ ε), where k is the maximum number of paths in S that are needed for composing any path in P, and d is the maximum number of times one edge appears in S. We also derive analogous results for approximate differential privacy in both the general case and for well-structured graphs. Beyond these main results, we also show that the paths between L-pairs of vertices in cycles admit (polylog(n), polylog(n))-summary sets with size poly(n), which implies tighter worst-case error guarantees for distance estimation.

Joint work with Tamalika Mukherjee, Jalaj Upadhyay, Hantao Yu, and Zongrui Zou

Speaker's Bio

Dr. Rachel Cummings is an Associate Professor of Industrial Engineering and Operations Research and (by courtesy) Computer Science at Columbia University, where she is also a member of the Data Science Institute and co-chairs the Cybersecurity Research Center. Before joining Columbia, she was an Assistant Professor of Industrial and Systems Engineering and (by courtesy) Computer Science at the Georgia Institute of Technology, and she received her Ph.D. in Computing and Mathematical Sciences from the California Institute of Technology. Her research interests lie primarily in data privacy, with connections to machine learning, algorithmic economics, optimization, statistics, and public policy. Dr. Cummings is the recipient of numerous awards including an NSF CAREER award, a DARPA Young Faculty Award, a DARPA Director’s Fellowship, an Early Career Impact Award, multiple industry research awards, a Provost’s Teaching Award, two doctoral dissertation awards, and Best Paper Awards at DISC 2014, CCS 2021, and SaTML 2023. Dr. Cummings also serves on the Independent Census Scientific Advisory Committee, the ACM U.S. Technology Policy Committee, the IEEE Standards Association, and the Future of Privacy Forum’s Advisory Board, and was a Fellow at the Center for Democracy & Technology.

Abstract

In many inventory systems, unmet demand is censored by the amount of inventory in stock, so the decision-maker observes sales rather than demand. This creates a fundamental challenge for data-driven inventory control. In this talk, I will discuss how censoring affects learning in both offline and online problems. I will first consider the offline setting and characterize when the optimal ordering decision can be consistently learned, as well as the fundamental optimality gap when it cannot. I will then turn to the online setting, where ordering different quantities reveals different amounts of information about demand. I will introduce an “information ordering” framework that captures when data collected under one policy can be used to evaluate another, and show how exploiting this structure leads to improved performance guarantees.

Speaker's Bio

Sean Sinclair is an Assistant Professor of Industrial Engineering and Management Sciences at Northwestern University, where he also serves as Director of Graduate Studies. His research develops reinforcement learning algorithms for operations problems, with recent interests in inventory control and scheduling.  He received his PhD from Cornell University and was previously a postdoctoral associate at MIT.  His work received the 2026 SIGMETRICS Best Paper award, and his dissertation received honorable mention for both the SIGMETRICS Dissertation and the George Dantzig Dissertation awards.

Abstract

We study the use of delay information in queueing systems from two perspectives: a macro system-level view and a micro behavioral view. Our approach utilizes a diverse methodological toolkit, ranging from heavy-traffic limits to controlled behavioral experiments with human subjects. By bridging these methodologies, we highlight robust insights that emerge across different modeling techniques. Specifically, we show that more information is not always better and that vague or binary signaling can paradoxically enhance performance. Finally, we address the gap between rigorous mathematical modeling and the nuances of human decision-making, offering a way forward to integrate both into a unified, behaviorally-aware framework for operational design.

Speaker's Bio

I am a professor at the School of Management of University College London, where I currently head the Operations & Technology group. I serve on the editorial boards of Management Science and Queueing Systems as an associate editor, and I serve as an Area co-Editor of the Operations and Supply Chains area at Operations Research. My research interests lie in service operations management.  I am especially interested in the operational management of queueing systems, from both mathematical and behavioral perspectives.

Abstract

TBD

Speaker's Bio

TBD

Abstract

TBD

Speaker's Bio

TBD

Abstract

TBD

Speaker's Bio

TBD

Abstract

TBD

Speaker's Bio

TBD