h-llpt-24
收藏资源简介:
h-llpt-24数据集由德国康斯坦茨大学和卢森堡科学技术研究所创建,用于评估机器学习原子间势能模型在高压氢液-液相变模拟中的性能。该数据集包含612个密度泛函理论分子动力学模拟的参考数据,涵盖不同温度和质量密度下的几何结构、能量、力和应力。数据集的创建过程包括在不同条件下进行分子动力学模拟,并从中提取训练和测试数据。该数据集主要应用于机器学习模型的性能评估,旨在解决高压氢相变模拟中的准确性和效率问题。
The h-llpt-24 dataset was developed by the University of Konstanz (Germany) and the Luxembourg Institute of Science and Technology, aiming to evaluate the performance of machine learning interatomic potential models in simulations of high-pressure hydrogen liquid-liquid phase transitions. This dataset includes 612 sets of reference data generated via density functional theory (DFT) molecular dynamics simulations, covering atomic geometries, energies, forces, and stresses under varying temperatures and mass densities. The dataset construction process involves performing molecular dynamics simulations under diverse conditions and extracting training and test datasets from the simulation outputs. It is primarily applied to performance evaluation of machine learning models, with the goal of resolving the accuracy and efficiency challenges in high-pressure hydrogen phase transition simulations.

- 1Hydrogen under Pressure as a Benchmark for Machine-Learning Interatomic Potentials德国康斯坦茨大学计算机与信息科学系,卢森堡科学技术研究所 · 2024年



