Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
DY: Fachverband Dynamik und Statistische Physik
DY 51: Machine Learning in Dynamics and Statistical Physics (joint session DY/SOE)
DY 51.4: Vortrag
Freitag, 9. September 2022, 10:45–11:00, H19
Exploring structure-property maps with kernel principal covariates regression — •Guillaume Fraux, Benjamin Helfrecht, Rose Cersonsky, and Michele Ceriotti — Institute of Materials, EPFL, Lausanne, Switzerland
Data analyses based on linear methods constitute the simplest, most robust, and transparent approaches to the automatic processing of large amounts of data for building supervised or unsupervised machine learning models. Principal covariates regression (PCovR) is an underappreciated method that interpolates between principal component analysis and linear regression and can be used conveniently to reveal structure-property relations in terms of simple-to-interpret, low-dimensional maps. We introduce a kernel version of PCovR (KPCovR), and demonstrate the performance of this approach in revealing and predicting structure-property relations in chemistry and materials science.
For large datasets, interactive exploration of the resulting map is a great tool to extract understanding. To this end, we introduce chemiscope, an open source software able to display and explore maps with hundred of thousands of points together with the corresponding molecular or crystal structure. Chemiscope is usable as an online tool, or locally through jupyter notebooks.