Kurzfassung

Mathematical models are frequently used to predict and analyse the behaviour of complex dynamic systems. In many areas like biology or economics, a major problem of model development is uncertain or incomplete qualitative knowledge about the system of interest. In this project we are developing statistical learning algorithms to combine models with data and to deal with this uncertainty. We have already published a first paper describing the dynamic elastic-net algorithm for estimating...Mathematical models are frequently used to predict and analyse the behaviour of complex dynamic systems. In many areas like biology or economics, a major problem of model development is uncertain or incomplete qualitative knowledge about the system of interest. In this project we are developing statistical learning algorithms to combine models with data and to deal with this uncertainty. We have already published a first paper describing the dynamic elastic-net algorithm for estimating dynamic model errors in differential equation models from data. The dynamic elastic-net is is also a methiod to estimate state variables which can not directly be measured, even when the underying model is incomplete or partially incorrect. We continue our work to extend this approach to larger systems and to quantify the uncertainty the state and model error estimates. » weiterlesen» einklappen