Sequential and adaptive learning under dependence and non-standard objective functions (A03)

• Subject Area: Mathematics
• Term: since 2017
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 318763901

The project addresses the question of how one can optimally collect data sequentially and adaptively with regard to a given objective function, where at each time step only limited information can be acquired about the underlying system, but the user can decide sequentially what is observed. In contrast to classical sequential learning and experimental design theory, the data is correlated in time and assumed to arise from an underlying stochastic process. We focus on the case in which the set of actions has a specific topology that can be exploited – for example, one induced by a graph or a smooth function.

• DFG Programme: Collaborative Research Centres
• Subproject of: SFB 1294: Data Assimilation – The Seamless Integration of Data and Models
• Applicant Institution: Universität Potsdam
• Project Heads: Professor Dr. Gilles Blanchard, until 6/2021; Professorin Dr. Alexandra Carpentier; Professorin Dr. Jana De Wiljes, since 7/2021