Dagstuhl Seminar 28081
Scheduling
( Feb 20 – Feb 25, 2028 )
Permalink
Please use the following short url to reference this page:
https://www.dagstuhl.de/28081
Organizers
- Franziska Eberle (TU Berlin, DE)
- Benjamin J. Moseley (Carnegie Mellon University - Pittsburgh, US)
- Viswanath Nagarajan (University of Michigan - Ann Arbor, US)
- Marc Uetz (University of Twente - Enschede, NL)
Contact
- Marsha Kleinbauer (for scientific matters)
- Jutka Gasiorowski (for administrative matters)
Scheduling is a central topic in algorithms and optimization, with applications in manufacturing, project planning, cloud computing, energy systems, and communication networks. Much of classic scheduling theory is built around worst-case models. In many modern applications, however, key inputs are uncertain: job sizes may be unknown, arrivals may be random, and the underlying distributions may themselves be only partially known or estimated from data.
At the same time, several neighboring areas have developed tools that have not yet been fully integrated into scheduling. Stochastic scheduling has produced strong combinatorial techniques, including mathematical programming relaxations, rounding methods, dynamic programming, concentration arguments, and hardness reductions. Stochastic combinatorial optimization has developed ideas ranging from prophet inequalities, secretary problems, online matching, to sample-based algorithms. Distributionally robust optimization and distributional learning offer ways to reason about settings in which the distribution is misspecified or must be learned from data.
This Dagstuhl Seminar will bring these communities together around a set of basic questions. How much distributional information does a scheduling algorithm actually need? Can good policies be designed from a few samples, or from partial information such as moments? What happens when the assumed distribution is inaccurate? How should scheduling algorithms change when they can learn from repeated observations?
We will study these questions through four closely connected themes: stochastic scheduling, combinatorial stochastic optimization, distributional robustness, and learning distributions over time. The aim is not only to bring new stochastic and learning tools into scheduling, but also to understand where techniques developed for scheduling can be useful in other discrete stochastic optimization problems. By putting these perspectives in direct contact, we hope to identify new models, techniques, and open problems for scheduling and discrete optimization under uncertainty.
Creative Commons BY 4.0
Franziska Eberle, Benjamin J. Moseley, Viswanath Nagarajan, and Marc Uetz
Related Seminars
- Dagstuhl Seminar 08071: Scheduling (2008-02-10 - 2008-02-15) (Details)
- Dagstuhl Seminar 10071: Scheduling (2010-02-14 - 2010-02-19) (Details)
- Dagstuhl Seminar 13111: Scheduling (2013-03-10 - 2013-03-15) (Details)
- Dagstuhl Seminar 16081: Scheduling (2016-02-21 - 2016-02-26) (Details)
- Dagstuhl Seminar 18101: Scheduling (2018-03-04 - 2018-03-09) (Details)
- Dagstuhl Seminar 20081: Scheduling (2020-02-16 - 2020-02-21) (Details)
- Dagstuhl Seminar 23061: Scheduling (2023-02-05 - 2023-02-10) (Details)
- Dagstuhl Seminar 25121: Scheduling (2025-03-16 - 2025-03-21) (Details)
Classification
- Data Structures and Algorithms
- Discrete Mathematics
Keywords
- scheduling
- algorithms and complexity
- stochastic optimization
- online learning
- distributionally robust optimization
