Collaborative Research: Framework: Machine Learning Materials Innovation Infrastructure

合作研究：框架：机器学习材料创新基础设施

• 批准号: 1931298
• 负责人: Dane Morgan
• 金额: $ 158.06万
• 依托单位: University of Wisconsin-Madison
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-10-01 至 2024-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1931298&HistoricalAwards=false
• 关键词: Collaborative; Research; Framework; Machine; Learning

Machine learning is rapidly changing our society, with computers recently gaining skills in many new tasks. These tasks range from understanding language to driving cars. Materials science and engineering is also being transformed. Many tasks are becoming increasingly accessible to machine learning algorithms. These range from predicting new data to analyzing images. Many basic machine learning algorithms are readily available. However the overall workflow involved in the application of machine learning for materials problems is still largely executed by hand. Getting results out is still done by traditional methods like publishing articles. There is an enormous opportunity to accelerate the growth and impact of machine learning in materials research. This requires improved cyberinfrastructure. This project will develop an approach to accelerate the entire machine learning workflow. Its output will include tools to easily develop datasets, manage model development, and output models. These will be reusable and reproducible for future use. This project will enable materials scientists and engineers to rapidly develop and deploy machine learning models. More importantly, the entire materials community will be able to quickly access these models. It will transform how we discover and develop advanced materials.The project will have three major technical components: (i) A MAterials Simulation Toolkit for Machine Learning (MAST-ML) with workflow tools that will enable local or cloud-based multistep, automated execution of complex machine learning data analysis and model training, codified best practices, increased access to machine learning methods for non-experts, and accelerated model development; (ii) The Foundry Materials Informatics Environment that will provide flexible, integrated, cloud-based management of machine learning materials science and engineering projects, from organizing data to developing models to disseminating results that are machine and human accessible and reproducible in ways that support a networked materials innovation ecosystem, (iii) Representative science applications of machine learning materials science and engineering projects that will support infrastructure development and promotion, as well as demonstrate best practices on state-of-the-art materials science and engineering problems. In addition to its impact on materials science and engineering, this project will develop students and young researchers with the interdisciplinary skills of machine learning and materials science and engineering, and promote these new ideas to the broader materials community. This award is jointly supported by the NSF Office of Advanced Cyberinfrastructure, and the Division of Materials Research within the NSF Directorate of Mathematical and Physical Sciences.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

机器学习正在迅速改变我们的社会，计算机最近在许多新任务中获得了技能。这些任务的范围从理解语言到驾驶汽车。材料科学和工程也在发生变化。机器学习算法越来越容易处理许多任务。这些范围从预测新数据到分析图像。许多基本的机器学习算法是现成的。然而，机器学习在材料问题中的应用所涉及的整体工作流程仍然主要由手工执行。获得结果仍然是通过传统的方法，如发表文章。这是一个巨大的机会，可以加速机器学习在材料研究中的发展和影响。这需要改善网络基础设施。该项目将开发一种方法来加速整个机器学习工作流程。其输出将包括轻松开发数据集、管理模型开发和输出模型的工具。这些将是可重复使用和可复制的，以供将来使用。该项目将使材料科学家和工程师能够快速开发和部署机器学习模型。更重要的是，整个材料社区将能够快速访问这些模型。它将改变我们发现和开发先进材料的方式。该项目将有三个主要技术组成部分：（i）机器学习材料模拟工具包（MAST-ML），其工作流程工具将使复杂的机器学习数据分析和模型培训能够在本地或基于云的多步骤自动执行，编纂最佳做法，增加非专家获得机器学习方法的机会，加快型号研制;（ii）铸造材料信息学环境，将为机器学习材料科学和工程项目提供灵活、集成、基于云的管理，从组织数据到开发模型，再到传播机器和人类可访问和可复制的结果，以支持网络化材料创新生态系统，（iii）机器学习材料科学和工程项目的代表性科学应用，这些项目将支持基础设施的发展和推广，并展示最先进的材料科学和工程问题的最佳做法。除了对材料科学和工程的影响外，该项目还将培养学生和年轻研究人员掌握机器学习和材料科学与工程的跨学科技能，并将这些新想法推广到更广泛的材料社区。该奖项由美国国家科学基金会高级网络基础设施办公室和美国国家科学基金会数学与物理科学理事会材料研究部共同支持。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

Deep learning object detection in materials science: Current state and future directions

• DOI: 10.1016/j.commatsci.2022.111527
• 发表时间: 2022-05-24
• 期刊: COMPUTATIONAL MATERIALS SCIENCE
• 影响因子: 3.3
• 作者: Jacobs, Ryan

HydroNet: Benchmark Tasks for Preserving Intermolecular Interactions and Structural Motifs in Predictive and Generative Models for Molecular Data

• 发表时间: 2020-11
• 期刊: ArXiv
• 作者: Sutanay Choudhury;Jenna A. Bilbrey;Logan T. Ward;S. Xantheas;Ian T Foster;Josef Heindel;B. Blaiszik

Multi defect detection and analysis of electron microscopy images with deep learning

• DOI: 10.1016/j.commatsci.2021.110576
• 发表时间: 2021-08
• 期刊: ArXiv
• 作者: Mingren Shen;Guanzhao Li;Dongxian Wu;Yuhan Liu;J. Greaves;Wei Hao;Nathaniel J. Krakauer;Leah Krudy;J. Perez;V. Sreenivasan;Bryan Sanchez;Oigimer Torres;Wei Li;K. Field;D. Morgan

Graph network based deep learning of bandgaps

基于图网络的带隙深度学习

• DOI: 10.1063/5.0066009
• 发表时间: 2021
• 期刊: The Journal of Chemical Physics
• 作者: Li, Xiang-Guo;Blaiszik, Ben;Schwarting, Marcus Emory;Jacobs, Ryan;Scourtas, Aristana;Schmidt, K. J.;Voyles, Paul M.;Morgan, Dane
• 通讯作者: Morgan, Dane

How machine learning is revolutionising materials science

机器学习如何彻底改变材料科学

• DOI: 10.33424/futurum377
• 发表时间: 2023
• 期刊: Futurum Careers
• 作者: Morgan, Dane;Jacobs, Ryan
• 通讯作者: Jacobs, Ryan