Data-driven complete basis set limit estimates from a minimal auxiliary basis

The dataset is based on the RHF dataset from GDB-BSIE and incorporates pcseg-0 energies, along with those for three versions of the Complementary Auxiliary Basis Set (CABS) for that basis set and one for cc-pVDZ. The details of each version are provided in the manuscript. Main FilesGDB-BSIE_CABS.tar.gz - Main dataset based on GDB-BSIE from Holm et al. [1], including CABS-related information. Delta_Learning.tar.gz- Δ-learning datasetsDirect_Learning.tar.gz- Direct-learning datasets. The learning tar files are organized according to: Delta/Direct Learning ├── minimal_RHF-F12-pcseg-0 ├── RHF-cc-pvdz ├── RHF-F12-cc-pvdz ├── RHF-F12-pcseg-0 ├── RHF-pcseg-0 ├── RHF-sz-dz └── small_RHF-F12-pcseg-0 Each folder describes the different trained basis sets.* Inside are the jsonl files for each one of the 4 descriptors to reproduce the learning curves of this work. In the Delta approach, there are additional results beyond those of the Global descriptor: For minimal_RHF-F12, pairwise representations are indicated by the suffix -pair For small_RHF-F12, pairwise representations indicated by the suffix -pair and a Local descriptor (marked as Local instead of Global) Molecular Representations - FCHL19 - MACE - SLATM - cMBDF * In the manuscript, energies are named as CABS, while here they are named F12. * The results are the average of 5 runs [1] Holm, S., Unzueta, P. A., Thompson, K. & Martínez, T. J. Single-Point Extrapolation to the Complete Basis Set Limit through Deep Learning. J. Chem. Theory Comput. 19, 4474–4483 (2023).