Significance of materials informatics and the development of new materials
Significance of materials informatics and the development of new materials
复制标题
材料信息学的意义和新材料的发展
DOI：
10.11470/jsaprev.220416
复制
发表时间：
2022
期刊：
影响因子：
--
通讯作者：
Mikiya Fujii
中科院分区：
文献类型：
--
作者：
竹内康基;加藤俊介;林高史;Mikiya Fujii
近年来，一个被称为材料信息学（MI）的领域引起了人们的注意[1，2]。MI是材料科学和信息科学的交叉领域。机器学习和深度学习技术经常应用于材料开发。它们是数字技术;到目前为止，还开发了其他数字技术，如第一性原理计算和分子动力学计算，但这两类技术之间存在差异。前者强调归纳点，而后者强调演绎点。后一派通过用计算机求解自然科学的控制方程来解释特定的现象。相反，在机器学习/深度学习中，准备一组数据，估计表示数据集的特征的统计模型，并对其他数据执行预测。在这里，我不相信自然科学的控制方程，这是指导原则，是必要的作出预测。相反，即使对于那些控制方程不明显的现象，也可以使用数据的统计模型进行预测。通常，这些解释不能完全满足自然科学家。然而，有一个优点，即可预测性是可能超出第一原理和分子动力学计算的应用。在本文中，我将回顾MI，讨论它的意义，并介绍了作者的小组已经参与了最近的研究的例子。
In recent years, a field called materials informatics (MI) has attracted attention [1, 2]. MI is an interdisciplinary field of materials science and information science. Machine learning and deep learning technologies are often applied to materials development. They are digital technologies; other digital technologies, such as first-principles calculations and molecular dynamics calculations, have been developed so far, but there are differences between the two groups. The former group emphasizes inductive points, whereas the latter group emphasizes deductive points. The latter group explains specific phenomena by solving natural science governing equations with a computer. In contrast, in machine learning/deep learning, a set of data is prepared, a statistical model representing the characteristics of the data set is estimated, and prediction is performed on other data. Here, I do not believe that the natural scientific governing equations, which are the guiding principles, are necessary for making predictions. Conversely, even for phenomena whose governing equations are not apparent, predictions are made using statistical models from data. Often, these explanations do not fully satisfy the natural scientist. However, there is an advantage that predictability is possible beyond the application of first-principles and molecular dynamics calculations. In this paper, I will review MI, discuss its significance, and introduce examples of recent studies in which the authors’ group has been involved.