QeMFi: A Multifidelity Dataset of Quantum Chemical Properties of Diverse Molecules

This is the QeMFi (Quantum chemistry MultiFidelity) dataset (previously CheMFi, name updated post peer review process) generated for the molecules of the WS22 database. Multifidelity machine learning (MFML) for quantum chemical properties involves building a composite model using various fidelities as opposed to a single fidelity. This dataset is presented to the community as a collection of multifidelity data for various quantum chemical properties for benchmarking of future MFML models. The README.md file contains more details about this dataset and use-cases.本数据集为QeMFi（Quantum chemistry MultiFidelity，原名称为CheMFi，经同行评审流程后完成更名）数据集，系针对WS22数据库中的分子生成而来。针对量子化学性质的多保真度机器学习（Multifidelity machine learning, MFML），其核心思路是采用多种保真度层级构建复合模型，而非仅依赖单一保真度。本数据集面向科研社区公开，收录了涵盖多种量子化学性质的多保真度数据，可用于未来MFML模型的基准测试。该数据集的详细信息与应用场景可参阅README.md文件。