Matbench
收藏资源简介:
Matbench是一个专门为无机固体材料性质预测设计的基准测试套件,由劳伦斯伯克利国家实验室创建。该数据集包含13个机器学习任务,涵盖了从312到132,752个样本的不同规模,数据来源于10个密度泛函理论衍生的和实验的源。任务包括预测光学、热学、电子、热力学、拉伸和弹性等性质。Matbench旨在通过一致的嵌套交叉验证方法评估回归和分类误差,适用于多种机械、电子和热力学材料性质。该数据集的应用领域广泛,旨在解决材料科学中的预测问题,推动材料信息学的发展。
Matbench is a benchmark suite dedicated to property prediction of inorganic solid materials, developed by Lawrence Berkeley National Laboratory. This dataset includes 13 machine learning tasks, spanning dataset sizes ranging from 312 to 132,752 samples, with data sourced from 10 density functional theory-derived and experimental sources. The tasks cover the prediction of optical, thermal, electronic, thermodynamic, tensile and elastic properties. Matbench aims to evaluate regression and classification errors through a consistent nested cross-validation method, and is suitable for a variety of mechanical, electronic and thermodynamic material properties. This dataset has broad application scenarios, aiming to solve prediction problems in materials science and promote the development of materials informatics.




