HDR IDEAS^2 Institute: Data-Driven Frameworks for Materials Discovery

HDR IDEAS^2 Institute：材料发现的数据驱动框架

• 批准号: 1934641
• 负责人: Samantha Daly
• 金额: $ 200万
• 依托单位: University of California-Santa Barbara
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-09-01 至 2022-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1934641&HistoricalAwards=false
• 关键词: HDR; IDEAS; Institute; Data; Driven

The discovery and development of new materials with unique properties and functionalities has revolutionized entire industries, including aviation, space, communication, biomedical, and automotive. Materials design has been traditionally experimentally and computationally intensive. However, advances in data-driven approaches, computational power, and experimental capabilities have created a tipping point for targeted and efficient materials design. This Harnessing the Data Revolution Institutes for Data-Intensive Research in Science and Engineering (HDR-I-DIRSE) Frameworks award supports conceptualization of an Institute to advance data-intensive research in Materials Science and Engineering. The IDEAS^2 (Integrated Data Environment for Accelerated Stochastic Science) Institute for Materials Discovery will provide a platform for the development of experimental and computational frameworks for materials advancement, that encourages collaboration and the sharing of data-driven approaches among research communities. The Data Science methods are intrinsically interoperable, and this program will engage diverse research communities in the collaborative development of large data frameworks that are applicable across a wide range of disciplines. The IDEAS^2 Institute will be structured to lower the barrier for domain scientists to work with data scientists through a variety of mechanisms including biannual "Teach the Teacher" workshops, an annual IDEAS^2 Symposium, visiting faculty positions at UCSB, and a range of other community engagement activities. Students working on this program will gain valuable multidisciplinary research and educational opportunities.First-principle calculations of thermodynamic and kinetic properties and information from microstructurally-based, high throughput models will be integrated into the design of data structures and the analyses of the developed techniques. The developed frameworks will be grounded in machine learning approaches that are fundamentally-based, computationally and statistically tractable, and incorporate domain knowledge and simulation results. The frameworks and data developed in the Institute - such as those to predict processing advancements from first principles, model these advancements in a high-throughput fashion, enable high-throughput experimentation, align the resulting experimental data (chemical, microstructure, deformation, etc.), and efficiently mine the resultant high-dimensional datasets - will be integrated with an open-source platform (BisQue) to facilitate both internal and external collaboration on their development for a broad range of materials applications. The computational infrastructure and parallelization of calculations through the BisQue platform enables the screening of very large datasets, with a hierarchical workflow requiring minimal software requirements (only a web browser is needed) and minimal domain knowledge of the user in modeling of materials. The focus of this program is on a research area with major and broad implications on numerous scientific and technological fields, and it also represents a unique training opportunity with acquired skills that will propel its graduates to the forefront of the emerging, critical field of data-driven science, as well as its many application areas within various scientific disciplines and high-tech industry sectors. This project is part of the National Science Foundation's Harnessing the Data Revolution (HDR) Big Idea activity and is co-funded by the Division of Civil, Mechanical and Manufacturing Innovation.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.

具有独特性能和功能的新材料的发现和开发已经彻底改变了整个行业，包括航空，航天，通信，生物医学和汽车。材料设计传统上是实验和计算密集型的。然而，数据驱动方法、计算能力和实验能力的进步为有针对性和高效的材料设计创造了一个临界点。“利用数据革命促进科学与工程数据密集型研究机构”（HDR-I-DIRSE）框架奖支持一个机构的概念化，以推进材料科学与工程领域的数据密集型研究。IDEAS^2（加速随机科学集成数据环境）材料发现研究所将为材料进步的实验和计算框架的发展提供一个平台，鼓励研究团体之间的合作和数据驱动方法的共享。数据科学方法本质上是可互操作的，该计划将使不同的研究团体参与到适用于广泛学科的大型数据框架的协作开发中。IDEAS^2研究所的结构将通过各种机制降低领域科学家与数据科学家合作的障碍，包括两年一次的“Teach The Teacher”研讨会、每年一次的IDEAS^2研讨会、访问UCSB的教师职位，以及一系列其他社区参与活动。参加本课程的学生将获得宝贵的多学科研究和教育机会。热力学和动力学性质的第一性原理计算以及基于微结构的高通量模型的信息将集成到数据结构的设计和已开发技术的分析中。开发的框架将以机器学习方法为基础，这些方法基于基础，可计算和统计处理，并结合领域知识和模拟结果。研究所开发的框架和数据-例如从第一性原理预测加工进步的框架和数据，以高通量的方式模拟这些进步，实现高通量实验，对齐所得实验数据（化学，微观结构，变形等）；并有效地挖掘由此产生的高维数据集-将与一个开源平台（BisQue）集成，以促进内部和外部合作，以开发广泛的材料应用。通过BisQue平台的计算基础设施和并行化计算使筛选非常大的数据集成为可能，分层工作流程需要最少的软件要求（只需要一个web浏览器）和用户在材料建模方面的最少领域知识。该计划的重点是对众多科学和技术领域具有重大和广泛影响的研究领域，它也代表了一个独特的培训机会，获得技能，将推动其毕业生进入新兴的关键数据驱动科学领域的前沿，以及在各种科学学科和高科技产业部门的许多应用领域。该项目是美国国家科学基金会“利用数据革命（HDR）大创意”活动的一部分，由民用、机械和制造创新部共同资助。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

Stage-wise Conservative Linear Bandits

• 发表时间: 2020-09
• 期刊: ArXiv
• 作者: Ahmadreza Moradipari;Christos Thrampoulidis;M. Alizadeh

Decentralized Multi-Agent Linear Bandits with Safety Constraints

具有安全约束的去中心化多智能体线性强盗

• 发表时间: 2021
• 期刊: Proceedings of the AAAI Conference on Artificial Intelligence
• 作者: Amani, S.;Thrampoulidis, C.
• 通讯作者: Thrampoulidis, C.

Towards Optimal Off-Policy Evaluation for Reinforcement Learning with Marginalized Importance Sampling

• 发表时间: 2019-06
• 期刊: Genetica
• 影响因子: 1.5
• 作者: Tengyang Xie;Yifei Ma;Yu-Xiang Wang

Mechanical Metrics of Virtual Polycrystals (MechMet)

虚拟多晶的力学指标 (MechMet)

• DOI: 10.1007/s40192-021-00206-7
• 发表时间: 2021
• 期刊: Integrating Materials and Manufacturing Innovation
• 影响因子: 3.3
• 作者: Dawson, Paul R.;Miller, Matthew P.;Pollock, Tresa M.;Wendorf, Joe;Mills, Leah H.;Stinville, Jean Charles;Charpagne, Marie Agathe;Echlin, McLean P.
• 通讯作者: Echlin, McLean P.

Safe Linear Thompson Sampling with Side Information

带有辅助信息的安全线性 Thompson 采样

• 发表时间: 2021
• 期刊: IEEE transactions on signal processing
• 影响因子: 5.4
• 作者: Moradipari, A.;Amani, S.;Alizadeh, M.;Thrampoulidis, C.
• 通讯作者: Thrampoulidis, C.