Work

Source data for “Learning when to use physical relationships in descriptor-based tabular machine learning”

Year: 2026

Type: dataset

Abstract: This dataset contains the source data of the article “Learning when to use physical relationships in descriptor-based tabular machine learning” (AI for Science). The study compares Direct, Physics-only, Residual-only, Ordinary stacking, PhysResStack and PhysicsGate models on 16 target–route endpoints from four published benchmarks of solid-state ionic conductors, thermoelectric materials, liquid metal batteries and photovoltaic devices. The deposit provides the seed-level evaluation metrics (Data S2), split identifiers (Data S1) and the values plotted in Figures 2–5 and S1–S3. Raw third-party datasets are not redistributed; their sources are listed in Table S1 of the Supplementary Information.

Source: ScienceDB

DOI: 10.57760/sciencedb.j00345.00006

Authors: Han Zhou, Shiwen Wang, Mengni Huo, Zhongyou Peng, Licheng Tan, Boryann Liaw, Hermann Tempel, Shicheng Yu, Yiwang Chen, Rüdiger-A. Eichel

Institutions: Forschungszentrum Jülich, RWTH Aachen University, Nanchang University, Gannan Normal University

Language: English

Topic: Machine Learning in Materials Science

Subfield: Materials Chemistry

Field: Materials Science

Domain: Physical Sciences

Open Access status: green

OpenAlex ID: W7225307898

Cites:

Cited by:

Related to: 10