Harnessing surrogate models for data-efficient predictive chemistry: descriptors <i>vs.</i> learned hidden representations
When data are scarce, surrogate models trained on QM descriptors help. We report that their hidden features, not the predicted descriptors, provide superior inputs for downstream predictive chemistry tasks.
Authors
- Thijs Stuyver (ORCID: https://orcid.org/0000-0002-8322-0572)
- Guan-Ming Chen (ORCID: https://orcid.org/0009-0004-2581-1712)
Institutions
- Centre National de la Recherche Scientifique (FR)
- Chimie ParisTech - PSL (FR)
- Université Paris Sciences et Lettres (FR)
Publication Details
- Journal
- Digital Discovery
- Published
- 2025-01-01
- DOI
- https://doi.org/10.1039/d5dd00256g
- Primary Topic
- Analytical Chemistry and Chromatography
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- Agence Nationale de la Recherche
- China Scholarship Council
- CHIST-ERA