DataScience@BI seminar with Daniel Heydecker
DataScience@bi invites 1851 Research Fellow Daniel Heydecker at Imperical College, London to give talk titled "Dynamical Large Deviations: In search of matching bounds".
- Starts:12:00, 3 December 2024
- Ends:13:00, 3 December 2024
- Location:BI - campus Oslo, B3 inner area - next to meeting room B3i-108 or Zoom
- Contact:Siri Johnsen (siri.johnsen@bi.no)
About
Daniel Heydecker is a 1851 Research Fellow at Imperial College, London, working on large deviations – between probability and PDE.
Abstract
Large deviations are a way of formulating the rare events of a model, which may encompass rare but catastrophic behaviour: The unlikelihood of different atypical behaviours is encoded into a so-called rate function. While bounding the rate function from below is a well-established theory, the success of accurately describing every possible atypical behaviour is much harder, and is tied to well-posedness of the evolution of the typical behaviour of the system. We discuss this in three models arising from physics, in which the correctness of the `guessed’ rate function is respectively proven, disproven, and unknown.