NSF Workshop: Machine Learning Hardware Breakthroughs Towards Green AI and Ubiquitous On-Device Intelligence. To be Held in November 2020.

本次研讨会旨在汇集来自学术界、工业界和政府机构的专家，讨论和确定机器学习硬件突破的前瞻性研究机会和挑战，以实现绿色人工智能和无处不在的机器学习驱动的智能。此外，讲习班将提供机会，形成不同学科的合作研究。为期三天的研讨会将于2020年11月举行。它将包括主题演讲，小组演讲和讨论，以及分组和总结会议，目的是将不同的研究社区聚集在一起，定义重要的研究挑战并促进机器学习硬件的突破。智力优势：对创新机器学习硬件的需求日益增长，这有可能带来数量级的硬件效率。然而，机器学习硬件的发展比机器学习算法的发展要慢得多。这是因为开发定制的机器学习加速器带来了巨大的挑战，因为（1）需要机器学习，微架构和物理芯片设计方面的跨学科知识，以及（2）由众多设计选择所带来的大设计空间，处理元件和存储器层次结构。该研讨会旨在汇集具有多元化专业知识的研究人员，讨论和确定机器学习硬件（包括电气和光学实现）的研究机会和挑战，以帮助实现人工智能（AI）和无处不在的机器学习驱动智能的突破。更广泛的影响：本次研讨会将邀请具有互补背景的研究人员，就潜在的研究挑战和实现机器学习硬件突破的方向提供不同的观点。可以产生新的想法来解决机器学习硬件的电气和光学实现的未来挑战。创新和商业化的机会，可以确定以下的研究思路。研究人员将有无与伦比的机会建立超越各自技术领域界限的学术和机构合作伙伴关系。最后，将鼓励来自代表性不足群体的研究人员以及早期职业研究人员的参与。研讨会的结果将通过研讨会报告传播。该奖项反映了NSF的法定使命，并被认为值得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来支持。

This workshop aims to bring together experts from academia, industry, and government agencies to discuss and identify visionary research opportunities and challenges for machine learning hardware breakthroughs towards green AI and ubiquitous machine learning powered intelligence. In addition, the workshop will provide opportunities to form collaborative research from different disciplines. This three-day workshop will be held virtually in November 2020. It will feature keynote speeches, panel presentations and discussions, as well as break-out and summary sessions, with the objective of bringing different research communities together, defining important research challenges and promoting machine learning hardware breakthroughs. Intellectual Merit: There has been a critical growing need for innovative machine learning hardware which has the potential to bring orders-of-magnitude hardware efficiency. However, the development of machine learning hardware is much slower than that of machine learning algorithms. This is because developing customized machine learning accelerators presents significant challenges due to (1) the need for cross-disciplinary knowledge in machine learning, micro-architecture, and physical chip design and (2) the large design space resulting from the numerous design choices of dataflows, processing elements, and memory hierarchy. This workshop aims to bring together researchers with a diversified set of expertise to discuss and identify research opportunities and challenges for machine learning hardware (both electrical and optical implementation) to assist in a road map for achieving breakthroughs in artificial intelligence (AI) and ubiquitous machine learning powered intelligence. Broader Impacts: This workshop will bring in researchers with complementary backgrounds to offer different perspectives on potential research challenges and directions for enabling machine learning hardware breakthroughs. Novel ideas could be generated to solve the future challenges for electrical and optical implementation of machine learning hardware. Innovation and commercialization opportunities may be identified following the research ideas. Researchers will have an unparalleled opportunity to build collaborative scholarly and institutional partnerships that transcend boundaries imposed by their respective technical areas. Finally, participation of researchers from underrepresented groups as well as early-career researchers will be encouraged. Results from the workshop will be disseminated through workshop reports.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.