Wang Zehua, Wang Li, Zhang Hao, Xu Hong, He Xiangming
Institute of Nuclear and New Energy Technology, Tsinghua University, Beijing, 100084, China.
Nano Converg. 2024 Feb 26;11(1):8. doi: 10.1186/s40580-024-00417-6.
Traditional methods for developing new materials are no longer sufficient to meet the needs of the human energy transition. Machine learning (ML) artificial intelligence (AI) and advancements have caused materials scientists to realize that using AI/ML to accelerate the development of new materials for batteries is a powerful potential tool. Although the use of certain fixed properties of materials as descriptors to act as a bridge between the two separate disciplines of AI and materials chemistry has been widely investigated, many of the descriptors lack universality and accuracy due to a lack of understanding of the mechanisms by which AI/ML operates. Therefore, understanding the underlying operational mechanisms and learning logic of AI/ML has become mandatory for materials scientists to develop more accurate descriptors. To address those challenges, this paper reviews previous work on AI, machine learning and materials descriptors and introduces the basic logic of AI and machine learning to help materials developers understand their operational mechanisms. Meanwhile, the paper also compares the accuracy of different descriptors and their advantages and disadvantages and highlights the great potential value of accurate descriptors in AI/machine learning applications for battery research, as well as the challenges of developing accurate material descriptors.
传统的新材料开发方法已不足以满足人类能源转型的需求。机器学习（ML）、人工智能（AI）及其进展使材料科学家认识到，利用人工智能/机器学习加速电池新材料的开发是一种极具潜力的工具。尽管将材料的某些固定属性用作描述符，以在人工智能和材料化学这两个独立学科之间架起桥梁的做法已得到广泛研究，但由于对人工智能/机器学习的运行机制缺乏了解，许多描述符缺乏通用性和准确性。因此，了解人工智能/机器学习的潜在运行机制和学习逻辑已成为材料科学家开发更准确描述符的必要条件。为应对这些挑战，本文回顾了此前关于人工智能、机器学习和材料描述符的工作，并介绍了人工智能和机器学习的基本逻辑，以帮助材料开发者理解其运行机制。同时，本文还比较了不同描述符的准确性及其优缺点，并强调了准确描述符在人工智能/机器学习应用于电池研究中的巨大潜在价值，以及开发准确材料描述符所面临的挑战。
