Objective

In the past twenty years the availability of vast dimensional data, typically referred to as big data, has given rise to exciting challenges in various fields of mathematics and computer sciences. The increasing need for getting a better understanding of such data in internet traffic, biology, genetics, and economics, has lead to a revolution in statistical and machine learning, optimisation and numerical analysis. Due to high dimensionality of modern statistical models, parameter estimation is a difficult task and statisticians typically investigate estimation methods under sparsity constraints. While an abundance of estimation algorithms is now available for high dimensional discrete models, a rigorous mathematical investigation of estimation problems for high dimensional continuous-time processes is completely undeveloped.

The aim of STAMFORD is to provide a concise statistical theory for estimation of high dimensional diffusions. Such high dimensional processes naturally appear in modelling particle interactions in physics, neural networks in biology or large portfolios in economics, just to name a few. The methodological part of the project will require development of novel advanced techniques in mathematical statistics and probability theory. In particular, new results will be needed in parametric and non-parametric statistics, and high dimensional probability, that are reaching far beyond the state-of-the-art. Hence, a successful outcome of STAMFORD will not only have a tremendous impact on statistical inference for continuous-time models in natural and applied sciences, but also strongly influence the field of high dimensional statistics and probability.