Fachbereich Mathematik

Machine Learning in classical DFT

Alessandro Simon, University of Tübingen

In recent years Machine Learning techniques have made great advances in
fields like Computer Vision and Natural Language Processing. While there
are also impressive results in the emerging field of Scientific Discovery they
are often highly tuned to the individual task like protein folding.
The more general field of Symbolic Regression (SR) aims at finding an _analytical_
expression describing the relationship of given input and output data. This can not only
be used to do "classical" data analysis, but also to find underlying equations of
physical systems by using experimental or simulation data.
Our discussion will focus on using the Equation Learner architecture (EQL) for finding an
approximate energy functional for a given classical many-body system. However, it can be easily
adapted to other problems.