遇见数据集

Data and scripts for "Δ-HGPR: A Hierarchical Gaussian Process Framework for Active-Learning Δ-Potentials"

收藏
Zenodo2026-07-28 更新2026-08-01 收录
官方服务:

资源简介:

Data, input/output files, and analysis scripts supporting the article"Δ-HGPR: A Hierarchical Gaussian Process Framework for Active-LearningΔ-Potentials". The archive covers three validation systems — a water toy force field,malonaldehyde, and the Zundel cation H5O2+ — and includes, for everycombination of base calculator (PM7, ANI-1ccx, MACE-MH-1) and random seedreported in the article: NTChem input and output files for theωB97X/6-31G(d) reference calculations, production molecular-dynamicstrajectories, trained Δ-HGPR model states with the fitted coefficients ofthe pair and many-body terms, free-energy-profile and uncertainty data,and the run logs of the active-learning cycles. Reference conditions(DFT-only and base-calculator-only) are included alongside the Δ-HGPRresults. Scripts for running the calculations, driving the active-learningcycle, and aggregating and analyzing the trajectories are also provided. Files:- water.tar.gz : water toy force field- malonaldehyde.tar.gz : malonaldehyde- H5O2p.tar.gz : Zundel cation H5O2+- README.md : directory structure and file-type description Reference electronic-structure calculations were performed with NTChem,which is available from the authors as an executable binary under theapplicable RIKEN license terms and is installed at supercomputer centersin Japan.

提供机构:
Zenodo
创建时间:
2026-07-28
二维码
社区交流群
二维码
科研交流群
商业服务