From Data to Robust Solutions in Optimisation Problems Event

Event details

Real-world optimisation problems usually suffer under uncertainty. This can be ranging from uncertain weather forecasts, over uncertain traffic demand, to minute measurements and calculation errors. But even a small error in the problem data can lead to a large difference in the resulting solution of an optimisation problem. In this talk I will talk about robust optimisation approaches to uncertainty, where only past observations are given. While most approaches in the literature assume that a "good", closed-form description of the uncertainty is available, the research into practical, data-driven uncertainty sets is still fairly young. I will cover approaches to reduce the number of scenarios, and talk about experiences with real-world shortest path data.