IUCRC Proposal Phase 1 Duke: Center for Alternative Sustainable and Intelligent Computing (ASIC)

IUCRC 提案第一阶段杜克大学：替代可持续和智能计算中心 (ASIC)

• 批准号: 1822085
• 负责人: Yiran Chen
• 金额: $ 75万
• 依托单位: Duke University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2018
• 资助国家: 美国
• 起止时间: 2018-09-15 至 2024-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1822085&HistoricalAwards=false
• 关键词: IUCRC; Proposal; Phase; Duke; Center

Neuromorphic circuits, which mimic the human brain's structure and processing ability on silicon, promise to enable ultralow power, high performance computing. Various key breakthroughs have emerged from exponential advances in computing technology and understanding brains, with increasing interest from governments, research institutes, and business entities. Nonetheless, existing computing architectures can hardly demonstrate universal advantages in a large variety of neuromorphic computing tasks. To address this challenge, this project will build the Center of Alternative Sustainable and Intelligent Computing (ASIC). ASIC is a collaborative industry-university partnership that joins leading industry members with academic researchers to conduct multidisciplinary science and research to enable high performance and energy efficient machine intelligence through novel circuits, architecture, and computation models.The Center of Alternative Sustainable and Intelligent Computing (ASIC) focuses on alternative computing platforms for cognitive applications, which are in general difficult to be efficiently performed on conventional von Neumann architecture. The ASIC Duke University (ASIC-Duke) Site will serve as the lead Site for ASIC and will lead the coordination and management of the Center. Research at the ASIC Duke Site will focus on industry-driven hardware innovations, especially infrastructure and implementation. The establishment of ASIC is expected to accelerate the development of new computing diagrams in industry sectors and expedites technology transfer from research discoveries to practical solutions.ASIC research will have broad impact on the computer and IT industries by introducing alternative computing platforms that complement the conventional computing platforms built on von Neumann architectures in the short term, and position alternative sustainable and intelligent computing equivalents important in long term. The knowledge and expertise developed via ASIC will effectively enable the predicted boom in cognitive applications and their urgent demands for hardware and software support. The talent and expertise of center PIs and the industry partners are integrated to enhance the competence of the center members and accelerate technology transfer.ASIC research may involve academic and industrial data along with public domain data. This data and the resulting research outputs will be maintained using appropriate best practices for each type of data for a period of three years after the closing of ASIC. A central repository, suitably tagged for appropriate referencing and documentation, will be set up at https://asic.pratt.duke.edu for maintaining the acquired and generated data from center projects. Access to all models and project results will be stored online and made available for downloading in near real time to respond to approved user requests.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.

神经形态电路在硅上模仿人脑的结构和处理能力，有望实现超低功率、高性能的计算。随着政府、研究机构和商业实体的兴趣与日俱增，计算技术和对大脑的理解取得了指数级的进步，出现了各种关键突破。然而，现有的计算体系结构很难在大量的神经形态计算任务中显示出普遍的优势。为了应对这一挑战，该项目将建立替代可持续和智能计算中心(ASIC)。ASIC是一种产学合作伙伴关系，将业界领先成员与学术研究人员结合起来，进行多学科的科学和研究，通过新的电路、架构和计算模型实现高性能和高能效的机器智能。替代可持续和智能计算中心(ASIC)专注于认知应用的替代计算平台，这些平台通常很难在传统的冯·诺伊曼架构上高效执行。ASIC杜克大学(ASIC-Duke)站点将作为ASIC的牵头站点，并将领导该中心的协调和管理。ASIC Duke网站的研究将专注于行业驱动的硬件创新，特别是基础设施和实施。ASIC的建立预计将加快工业领域新计算图表的开发，并加快技术从研究发现到实用解决方案的转移。ASIC研究将在短期内引入替代计算平台，补充基于冯·诺伊曼架构的传统计算平台，并在长期内定位重要的替代可持续和智能计算等价物，从而对计算机和IT行业产生广泛影响。通过ASIC开发的知识和专业技能将有效地支持认知应用的预期热潮及其对硬件和软件支持的迫切需求。中心PI和行业合作伙伴的人才和专业知识被整合起来，以增强中心成员的能力并加速技术转移。ASIC研究可能涉及学术和行业数据以及公共领域数据。在ASIC关闭后的三年内，这些数据和由此产生的研究产出将使用每类数据的适当最佳做法加以保存。Https://asic.pratt.duke.edu将建立一个中央储存库，适当标记以供适当的参考和记录，以维护从中心项目获得和生成的数据。对所有模型和项目结果的访问将存储在网上，并近乎实时地提供下载，以响应批准的用户请求。该奖项反映了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

MVStylizer: an efficient edge-assisted video photorealistic style transfer system for mobile phones

• DOI: 10.1145/3397166.3409140
• 发表时间: 2020-05
• 期刊: Proceedings of the Twenty-First International Symposium on Theory, Algorithmic Foundations, and Protocol Design for Mobile Networks and Mobile Computing
• 作者: Ang Li;Chunpeng Wu;Yiran Chen;Bin Ni

Learning Low-rank Deep Neural Networks via Singular Vector Orthogonality Regularization and Singular Value Sparsification

• DOI: 10.1109/cvprw50498.2020.00347
• 发表时间: 2020-04
• 期刊: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)
• 作者: Huanrui Yang;Minxue Tang;W. Wen;Feng Yan;Daniel Hu;Ang Li;H. Li;Yiran Chen

ESCALATE: Boosting the Efficiency of Sparse CNN Accelerator with Kernel Decomposition

• DOI: 10.1145/3466752.3480043
• 发表时间: 2021-10
• 期刊: MICRO-54: 54th Annual IEEE/ACM International Symposium on Microarchitecture
• 作者: Shiyu Li;Edward Hanson;Xuehai Qian;H. Li;Yiran Chen

How to Obtain and Run Light and Efficient Deep Learning Networks

• DOI: 10.1109/iccad45719.2019.8942106
• 发表时间: 2019-11
• 期刊: 2019 IEEE/ACM International Conference on Computer-Aided Design (ICCAD)
• 作者: Fan Chen;W. Wen;Linghao Song;Jingchi Zhang;H. Li;Yiran Chen

Soteria: Provable Defense against Privacy Leakage in Federated Learning from Representation Perspective

• DOI: 10.1109/cvpr46437.2021.00919
• 发表时间: 2020-12
• 期刊: 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
• 作者: Jingwei Sun;Ang Li;Binghui Wang;Huanrui Yang;Hai Li;Yiran Chen