Theory, algorithms, and applications for data modelling
An academic at the University of Southampton’s School of Electronics and Computer Science has just received funding to develop new approaches to data modelling and model-free data processing.
Dr Ivan Markovsky, who has devoted most of his career to numerical methods for data modelling has been awarded a €783,000 Starting Grant from the European Research Council (ERC) to develop his approach further. The (ERC) Starting Grants are awarded to excellent early career researchers to enable them to devote their time fully to research.
Dr Markovsky’s research has the unique aim of unifying and simplifying the growing number of data modelling methods and making them more easily applicable. “Data modelling is the only problem that is common to all areas of science and engineering,” said Dr Markovsky. “Moreover, models are often the bottleneck in applications. However, current data modelling knowledge is rather fragmented and repetitive.”
The central concept of his work is a mathematical problem, called structured low-rank approximation, which includes many applications and existing methods. “It is amazing that a diverse list of applications can be formulated and solved as a single core mathematical problem,” said Dr Markovsky.
A high-gain, high-risk objective of the project is model-free data processing, which bypasses the modelling stage and goes straight to the final goal, thus tackling the problem as a whole.
“Our research suggests ways to merge data modelling with model-based data processing, which allows us to skip the modelling step. This approach has been tried before but it is not efficient yet and needs more work,” Dr Markovsky added.
If successful, the research will have impact on applications in acoustics, biomedical signal processing, and bioinformatics.