Lunch colloquium Dr. Dimitris Boskos (assistant professor at DDC)

03 March 2021 12:30 till 13:30 - Location: online by Zoom - By: DCSC

"Distributionally Robust Data-Driven Uncertainty Quantification from Dynamic Measurements"

There is a plethora of control and optimization problems that entail dynamic random elements with an unknown probability distribution. At the same time, the designer seeks to make inferences about them using a limited amount of data, that may only reveal partial-state information of the process, frequently also corrupted by noise. Motivated by recent advances in distributionally robust optimization, in this talk, we explore the use of such measurements to construct Wasserstein ambiguity sets that track the evolution of the unknown probability distributions. We use tools from state estimation and uncertainty quantification to combine the information from the data with the known dynamic model and provide robust uncertainty descriptions for reliable decisions. The ambiguity sets are accompanied by quantifiable guarantees of containing the true distributions while their construction is based on first-principles assumptions like the classes where the unknown distributions of random initial conditions, parameters, and noise elements belong.

Join Zoom meeting:

Meeting ID: 976 9987 6999
Passcode: 234175