Dr. Oleksandra Kukharenko
Oleksandra Kukharenko obtained her PhD in mathematics at the Department of Cybernetics, Taras Shevchenko National University of Kyiv, Ukraine. In 2014 she joined the University of Konstanz, Germany (group of Prof. Christine Peter), first as a bridging fellow and than as ZIF-Marie Curie postdoctoral fellow, where she started to work on the analysis of the dynamics of protein conformations. Further she extended her research by looking at multi-body, multiscale problems as a Carl Zeiss fellow. In 2020 she became a group leader in the theory department at the Max Planck Institute for Polymer Research. In her research activity she combines molecular dynamics simulations with novel techniques of statistical analysis and machine learning to characterize the conformational phase space, to estimate kinetic and thermodynamical properties of soft matter systems.
Improving and extending nonparametric regression techniques
We use novel techniques of statistical/machine learning (ML) in order to address some of the problems that arise in the field of computational chemistry. We focus on a particularly promising ML approach, which combines the computational power of dynamical systems with an easy-to-implement supervised learning scheme. With this methodology we tackle several crucial problems that theoretical/computational chemistry is facing in the context of polymer studies. More specifically, we use the information processing ability of certain dynamical systems to reconstruct, explain, and predict some static and dynamic features of molecular systems of interest.
Feature importance analysis
Usually not all input data contains an important information for accurate prediction/classification results. Inferring the importance of the input features can help the physical interpretation of the obtained results, as well as is of great importance in solving the problem of ill conditioning of the regression models due to the high dimensionality of input data contrasting to the small number of samples. We use feature importance analysis for improving transparency of nonparametric prediction models and to improve the stability of the developed models.