HDR Institute: Institute for Data Driven Dynamical Design

HDR 研究所：数据驱动动态设计研究所

• 批准号: 2118201
• 负责人: Eric Toberer
• 金额: $ 1554.07万
• 依托单位: Colorado School of Mines
• 依托单位国家: 美国
• 项目类别: Cooperative Agreement
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-10-01 至 2026-09-30
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2118201&HistoricalAwards=false
• 关键词: HDR; Institute; Data; Driven; Dynamical

From molecules to robots, designing for dynamics has common theoretical underpinnings despite differences in length and time scale. However, such research is often overwhelmed by the high dimensional design space. The Institute for Data-Driven Dynamical Design addresses the challenge of prediction of dynamical processes in materials, including ion and molecular transport, catalytic pathways, and phase transformations in metamaterials, with a focus on discovering fundamentally new mechanisms and pathways. This research represents a paradigm shift from traditional material efforts involving incremental improvements in ground-state and steady-state properties. Developments in the data sciences target (i) strategies for encoding complex structures and mechanistic pathways for machine intelligence, (ii) new predictive capabilities for evolving systems, and (iii) advances in visualization and integrating machine and human expertise. Fueling these data science developments are large-scale simulations of dynamical processes across high dimensional design spaces. Experimental validation of these large-scale simulations addresses both end-product prediction and mechanistic pathways therein. The Institute's data science innovations may advance fields both within and beyond STEM involving complex time-evolving systems including molecular biology, atmospheric science, geophysics, and physical cosmology. The Institute seeks to grow and unite the dispersed data-driven design community. Long-term growth is sought through outreach activities involving (i) high school coding schools, (ii) undergraduate involvement in data-rich research, and (iii) a post-baccalaureate bridge program that introduces students to data sciences and motivate them to pursue higher degrees. Data-driven design community activities include (i) interdisciplinary summer schools and workshops, (ii) a Fellows program to collaboratively grow and disseminate the Institute’s developments, and (iii) dedicated efforts to create open-source software for the design community. Throughout these efforts, the Institute actively seeks to recruit, retain, and graduate a diverse array of students in STEM. This virtual Institute seeks to design complex dynamical materials and structures through the union of machine and human intelligence. To learn dynamical behavior and ultimately discover new mechanisms, three core data science needs are addressed: (i) new representations and learning architectures that capture and encode the spatial arrangement, interactions, and temporal evolution of complex materials and geometrical structures, (ii) efficient exploration of high dimensional, time-dependent design spaces, and (iii) new visual analytics tools to quantitatively incorporate human-in-the-loop design feedback. Advances in each of these areas form a virtuous cycle that accelerates discovery of new materials, driven by new mechanisms. This Institute converges an interdisciplinary team focused on four design spaces at their `tipping point', where large quantities of dynamical data can be readily created: (i) crystalline solids with tailored ion transport for fuel cells and batteries, (ii) pressure-sensitive metamaterials for robotics, (iii) light driven catalytic reactions for chemical production, and (iv) synthesis and assembly of porous frameworks for chemical separations. These four areas are testbeds for cyberinfrastructure development for the broader scientific community. Interwoven throughout these activities are dedicated activities to build a new generation of STEM talent at the intersection of data science and the physical sciences/engineering and to broaden participation in STEM through targeted outreach.This project is part of the National Science Foundation's Big Idea activities in Harnessing the Data Revolution (HDR). The award by the Office of Advanced Cyberinfrastructure is jointly supported by the Divisions of Chemistry, Materials Research, and Mathematical Sciences within the NSF Directorate for 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.

从分子到机器人，动力学设计有着共同的理论基础，尽管在长度和时间尺度上有所不同。然而，这种研究往往被高维的设计空间淹没了。数据驱动动态设计研究所致力于预测材料中的动态过程，包括离子和分子传输、催化路径和超材料中的相变，重点是从根本上发现新的机制和路径。这项研究代表了从传统物质努力的范式转变，涉及基态和稳态性质的渐进式改进。数据科学的发展目标是(I)为机器智能编码复杂结构和机械路径的战略，(Ii)进化系统的新预测能力，以及(Iii)可视化和整合机器和人类专业知识的进展。推动这些数据科学发展的是对高维设计空间动态过程的大规模模拟。这些大规模模拟的实验验证既涉及最终产品预测，也涉及其中的机制路径。该研究所的数据科学创新可能会推动STEM内外涉及复杂的时间演化系统的领域，包括分子生物学、大气科学、地球物理和物理宇宙学。该研究所寻求发展和联合分散的数据驱动型设计社区。通过(I)高中编码学校，(Ii)本科生参与数据丰富的研究，以及(Iii)向学生介绍数据科学并激励他们攻读更高学位的毕业后桥梁计划，寻求长期增长。数据驱动的设计社区活动包括(I)跨学科暑期学校和研讨会，(Ii)研究员计划，以合作发展和传播研究所的发展，以及(Iii)致力于为设计社区创建开放源码软件。通过这些努力，该研究所积极寻求在STEM招收、留住和毕业一批多样化的学生。这个虚拟研究所寻求通过机器和人类智能的结合来设计复杂的动态材料和结构。为了学习动态行为并最终发现新的机制，需要解决三个核心数据科学需求：(I)捕捉和编码复杂材料和几何结构的空间排列、相互作用和时间演变的新表示和学习架构，(Ii)高效探索高维、依赖时间的设计空间，以及(Iii)新的视觉分析工具，以定量地结合人在回路中的设计反馈。这些领域的进展形成了一个良性循环，在新机制的推动下，加速了新材料的发现。该研究所汇集了一个跨学科小组，重点放在四个处于“临界点”的设计空间，在这些空间中，可以容易地创建大量动态数据：(1)用于燃料电池和电池的具有量身定制的离子传输的晶体固体，(2)用于机器人的压敏超材料，(3)用于化学生产的光驱动催化反应，以及(4)用于化学分离的多孔框架的合成和组装。这四个领域是更广泛的科学界发展网络基础设施的试验田。这些活动交织在一起的是专门的活动，旨在培养数据科学和物理科学/工程交叉领域的新一代STEM人才，并通过有针对性的外展扩大对STEM的参与。该项目是国家科学基金会利用数据革命(HDR)的大创意活动的一部分。高级网络基础设施办公室的奖项由NSF数学和物理科学局内的化学、材料研究和数学科学部联合支持。这一奖项反映了NSF的法定使命，并通过使用基金会的智力优势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

Human–Computer Collaboration for Visual Analytics: an Agent‐based Framework

用于视觉分析的人机协作：基于代理的框架

• DOI: 10.1111/cgf.14823
• 发表时间: 2023
• 期刊: Computer Graphics Forum
• 影响因子: 2.5
• 作者: Monadjemi, Shayan;Guo, Mengtian;Gotz, David;Garnett, Roman;Ottley, Alvitta
• 通讯作者: Ottley, Alvitta

A Unified Comparison of User Modeling Techniques for Predicting Data Interaction and Detecting Exploration Bias

用于预测数据交互和检测探索偏差的用户建模技术的统一比较

• DOI: 10.1109/tvcg.2022.3209476
• 发表时间: 2023
• 期刊: IEEE Transactions on Visualization and Computer Graphics
• 影响因子: 5.2
• 作者: Ha, Sunwoo;Monadjemi, Shayan;Garnett, Roman;Ottley, Alvitta
• 通讯作者: Ottley, Alvitta

The Vendi Score: A Diversity Evaluation Metric for Machine Learning

• DOI: 10.48550/arxiv.2210.02410
• 发表时间: 2022-10
• 期刊: ArXiv
• 作者: Dan Friedman;Adji B. Dieng

Alloying-Induced Structural Transition in the Promising Thermoelectric Compound CaAgSb

• DOI: 10.1021/acs.chemmater.3c02621
• 发表时间: 2024-02
• 期刊: Chemistry of Materials
• 影响因子: 8.6
• 作者: A. Shawon;Weeam Guetari;Kamil M Ciesielski;Rachel Orenstein;Jiaxing Qu;Sevan Chanakian;Md. Towhidur Rahman;Elif Ertekin;Eric Toberer;Alexandra Zevalkink

Exploring Pre-Trained Language Models to Build Knowledge Graph for Metal-Organic Frameworks (MOFs)

• DOI: 10.1109/bigdata55660.2022.10020568
• 发表时间: 2022-12
• 期刊: 2022 IEEE International Conference on Big Data (Big Data)
• 作者: Yuan An;Jane Greenberg;Xiaohua Hu;Alexander Kalinowski;Xiao Fang;Xintong Zhao;Scott McClellan;F. Uribe-Romo;Kyle Langlois;Jacob Furst;Diego A. Gómez-Gualdrón;Fernando Fajardo-Rojas;Katherine Ardila;S. Saikin;Corey A. Harper;Ron Daniel