CyberTraining: Pilot: Employing Proper Orthogonal Decomposition (POD) and High-Performance Computing (HPC) in Advanced CI

网络培训：试点：在高级 CI 中采用适当的正交分解 (POD) 和高性能计算 (HPC)

• 批准号: 2118079
• 负责人: Daqing Hou
• 金额: $ 30万
• 依托单位: Clarkson University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-09-01 至 2024-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2118079&HistoricalAwards=false
• 关键词: CyberTraining; Pilot; Employing; Proper; Orthogonal

Proper Orthogonal Decomposition (POD) is a highly effective, data-driven learning algorithm for solving multi-dimensional Ordinary/Partial Differential Equations (ODEs/PDEs). However, POD is rarely covered in the typical graduate curriculum, and thus the nation is not fully leveraging this advanced algorithm in science and engineering research. To fill this void, this project will conduct a two-week online workshop for trainees in engineering and science-related disciplines and will integrate the developed instructional material into an existing graduate course on High-performance Computing (HPC). In doing so, this project provides for an educational ecosystem enabling computational and data-driven science for scientists and engineers. By training a diverse group of graduate students, post-docs, and faculty members in various disciplines, this project will help prepare the scientific workforce for advanced CI-enabled research, which will serve to enhance research productivity and enable researchers to effectively address complex societal problems. The project will help develop the national research workforce in areas of critical need through intensive, integrated instruction on open-source computing platforms to solve ODEs/PDEs by use of POD models that employ HPC. The training will incorporate team-based interdisciplinary projects based on data-driven POD learning algorithm for computationally intensive multiphysics simulation problems in various science and engineering disciplines. The workshop will provide trainees with intensive instruction on POD and related topics, including open source platforms to solve ODEs/PDEs and eigenvalue problems. Training culminates in a research project where trainees learn advanced computational tools and implementation of HPC skills. The project will broaden the access and adoption of advanced CI while integrating CI skills into existing curriculum models and fostering inter-disciplinary and inter-institutional research collaborations.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.

正确的正交分解（POD）是一种高效，数据驱动的学习算法，用于求解多维的普通/偏微分方程（ODES/PDES）。但是，POD很少在典型的研究生课程中涵盖，因此该国并未完全利用科学和工程研究中的先进算法。为了填补这一空白，该项目将为与工程和科学相关学科的学员进行为期两周的在线研讨会，并将开发的教学材料整合到现有的高性能计算研究生课程中。为此，该项目为科学家和工程师提供了一个教育生态系统，为科学和数据驱动的科学提供了教育生态系统。通过培训各种学科的各种研究生，博士后和教职员工，该项目将有助于为高级CI-Expair的研究准备科学劳动力，这将有助于提高研究生产力，并使研究人员能够有效解决复杂的社会问题。该项目将通过开源计算平台上的密集，集成的指导来帮助发展国家研究人员，以通过使用使用HPC的POD模型来解决ODE/PDE。该培训将基于基于数据驱动的POD学习算法的基于团队的跨学科项目，用于各种科学和工程学科中的计算密集型多物理模拟问题。该研讨会将为受训者提供有关POD和相关主题的密集指导，包括解决ODES/PDES和EIGENVALUE问题的开源平台。培训在一个研究项目中达到顶点，在该项目中，学员学习了高级计算工具和HPC技能的实施。该项目将扩大和采用高级CI的访问和采用，同时将CI技能整合到现有的课程模型中，并促进跨学科和机构间研究合作。该奖项反映了NSF的法定任务，并被认为是值得通过基金会的知识分子优点和更广泛影响的审查标准来通过评估来获得支持的。

项目成果

"Schrödinger Equation Solver Based on Data-Driven Physics-Informed Generic Building Blocks"

“基于数据驱动的物理信息通用构建块的薛定谔方程求解器”

• 发表时间: 2023
• 期刊: 2023.
• 作者: Veresko, Martin;Cheng, Ming-C.
• 通讯作者: Cheng, Ming-C.

An effective physics simulation methodology based on a data-driven learning algorithm

• DOI: 10.1145/3539781.3539799
• 发表时间: 2022-01-01
• 期刊: PASC '22: Proceedings of the Platform for Advanced Scientific Computing Conference
• 作者: Jiang, Lin;Veresko, Martin;Cheng, Ming-C
• 通讯作者: Cheng, Ming-C