Regensburg 2022 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

O: Fachverband Oberflächenphysik

O 58: Poster Wednesday: New Methods and Developments, Frontiers of Electronic Structure Theory

O 58.9: Poster

Mittwoch, 7. September 2022, 18:00–20:00, P4

Unsupervised regression-based measures for applications on atomistic features — •Alexander Goscinski¹, Guillaume Fraux¹, Giulio Imbalzano¹, Félix Musil^1,2, Sergey Pozdnyakow¹, and Michele Ceriotti¹ — ¹Laboratory of Computational Science and Modeling, Institute of Materials, École Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland — ²National Center for Computational Design and Discovery of Novel Materials (MARVEL), Lausanne, Switzerland

The quality of the features as input for a machine learning model is a crucial factor for the prediction quality and the computational efficiency. Commonly, to assess the quality of features, they are compared by benchmarking the regression performance on several properties. Complementary to such a quality assessment, this work presents certain measures for direct feature-to-feature comparisons without the need of a target property. These measures are used to quantify the capacity of features representing geometrical space in atomistic applications and derive an understanding of the information encoded in features.