活动名称：
【青年科学半月谈】机器学习预测材料性质：难点与改进策略
报告嘉宾：
龚盛 Massachusetts Institute of Technology
| 蔻享学术直播间
|
龚盛
Massachusetts Institute of Technology
Sheng Gong is a postdoc and recent graduate from Prof. Jeffrey C. Grossman’s group at Department of Materials Science and Engineering, Massachusetts Institute of Technology. He is interested in application of machine learning in materials science. His past and ongoing projects include study and design of graph neural networks for prediction of scalar and volumetric materials properties, application of information transfer strategy to improve prediction of materials properties with limited data, optimization of experimental conditions by Bayesian Optimization, and study and design of energy materials by machine learning and atomistic modeling.
Despite the widespread applications of machine learning models in materials science, in many cases the performance of machine learning models is not sufficiently accurate enough to meet the needs of materials design. In this presentation, we propose and apply a series of strategies to exam and improve uponthe performance of machine learning models for specific materials problems. First, we exam whether current deep representation learning models for atomistic systems can capture all human knowledge of crystal structures, and find that current graph neural networks can capture knowledge of local atomic environments but cannot capture periodicity of crystalline materials. As an initial solution, we hybridize human knowledge with deep representation learning models, and find that the hybridization can lead to improvement for predicting materials properties, especially vibrational properties. Then, for situations where the datasets of target materials properties are small while there are large relevant materials datasets, we propose to use transfer learning and multi-fidelity learning to transfer information between the large and small datasets to facilitate the learning of target properties. We use experimentally measured formation enthalpy and lattice thermal conductivity as case studies to exam the usefulness of information transfer and understand where and why information transfer helps. The machine learning models introduced in this presentation not only deepen human understanding of where and how machine learning can be used to facilitate materials development, but also lead to the discovery of new materials systems and new insights, such as new candidate thermoelectric materials and new insights into the evaluation of the stability of materials.
青年科学半月谈聚焦基础科学相关前沿研究，是促进海内外交叉学科青年学者学术交流的讨论会。在疫情流行的大背景下，我们旨在建立并提供一个自由、开放并充满活力的在线交流平台，以促进学科发展。欢迎广大的青年科学工作者（特别是在读研究生）来参加、交流、报告！
论坛特色
论坛欢迎在读研究生以及青年科研工作者来系统性地分享自己的工作。区别于一般的会议报告，我们在报告时长和篇幅上没有太多限制，并希望通过该论坛可以看到青年科研工作者的“第一视角”，进而展开更加有效且深入的讨论。（论坛无商业广告、无需提前报名注册、即点即听、后有回放可供反复学习交流、报告视频有正式doi索引）
蔻享学术平台，国内领先的一站式科学资源共享平台，依托国内外一流科研院所、高等院校和企业的科研力量，聚焦前沿科学，以优化科研创新环境、传播和服务科学、促进学科交叉融合为宗旨，打造优质学术资源的共享数据平台。
点击阅读原文，查看更多精彩报告！
