IBM Research - Ireland Internship Project: Identification of dynamical systems from sparse data sets - overview
In many engineering problems, one is interested in designing closed-loop control systems to impose a desired behaviour for a plant of interest. Typically, the control algorithms are model-based, i.e. rely on an abstraction of the physical system that needs to be controlled. Unfortunately, in many applications, the parameters of the model and the model structure are unknown and the datasets that could be used to identify the model are sparse.
Two typical examples of applications where sparsity of data makes it hard to identify the models of interest are water management systems (and, in general, chemical plants) and systems biology.
This internship is focused on the fundamental research problem of devising novel system identification techniques that allow to retrieve the parameters of a system of interest, given a sparse dataset.
- background in applied probability (with experience on e.g. multi-armed bandit problems);
- familiarity with Bayesian and maximum likelihood estimation methods;
- proficiency with Python.
- The proposed research is interdisciplinary and combines in a novel way concepts from control theory, big data and AI.