Chairs: Nikolas Nüsken, Sebastian Reich, Carsten Hartman, Nikki Vercauteren

Filtering, parameter estimation, data assimilation or smoothing of time series are inverse problems that can be rephrased as optimisation tasks. For multiscale systems, the oscillatory nature of the underlying forward problem creates additional difficulties, in addition to the general difficulty of solving inverse problems that are typically ill-posed. A common approach to deal with multiscale effects is to replace the original dynamics with a reduced model that is used to solve an inverse problem.

The mission of this minisymposium is twofold: (1) to discuss when such an approach is mathematically appropriate and algorithmically functional, (2) to discuss the application of surrogate modelling approaches based on novel optimisation and machine learning algorithms. It features contributions from both theorists and practitioners from various fields of science, mathematics and statistics.

Find abstracts for all talks here or linked individually below.

Continuation on Wednesday