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

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 2: Machine Learning Prediction and Optimization Tasks

AKPIK 2.4: Vortrag

Dienstag, 18. März 2025, 10:15–10:30, H5

Data nexus vista: A research assistance framework for physics-informed descriptor engineering — •Kanchan Sarkar and Axel Gross — Institute of Theoretical Chemistry, Ulm University, 89069 Ulm, Germany

In today's fast-paced world, accurate modeling and prediction of complex systems are more crucial than ever, particularly for advancing modern technologies. In energy storage, for example, the design of materials with tailored properties for sustainable, next-generation devices is a critical goal. However, the vast materials space and intricate nature of materials data poses serious challenges for experimental and computational explorations, often likened to searching for a needle in a haystack. To address this, we present "Data Nexus Vista," a flexible framework for descriptor engineering. The framework offers four core functionalities: (1) adaptive preprocessing protocols for defining physically meaningful data transformations; (2) a feature selection interface that incorporates physics-based rules for improved interpretability; (3) diverse correlation analysis tools to uncover underlying physical mechanisms; and (4) workflows for developing interpretable models that balance physical validity with statistical robustness. By enabling thorough exploration of data-physical phenomena connections, this user-guided, modular approach ensures scientific rigor while addressing uncertainties in complex systems. Moreover, its adaptable design allows seamless application across diverse domains, from energy storage, surface science to nonlinear optical properties, advancing both scientific discovery and practical innovation.

Keywords: Descriptor Engineering; Predictive Modeling; Correlation Analysis; Feature Selection