"Identifiability and structural inference for high-dimensional diffusion matrices"

We will develop theory and an estimation methodology for the analysis of sparsely parametrised high-dimensional diffusion matrices, including in particular identifiability issues. This requires the combination of adaptive smoothing techniques and sparsity inducing penalization methods and results in a challenging simultaneous adaptation problem. The analysis is supposed to provide more fundamental insight even for more classical situations for independent and identically distributed data.