RCN-SC: Research Coordination Network for Design and Testing of Neuromorphic Integrated Circuits

RCN-SC：神经形态集成电路设计和测试的研究协调网络

• 批准号: 2332166
• 负责人: Shantanu Chakrabartty
• 金额: $ 90万
• 依托单位: Washington University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-11-01 至 2026-10-31
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2332166&HistoricalAwards=false
• 关键词: RCN; SC; Research; Coordination; Network

The project is facilitating a Neuromorphic Integrated Circuits Education (NICE) research coordination network (RCN) that will lower the barrier of entry for students and researchers interested in the design, fabrication and testing of neuromorphic integrated circuits (ICs). As artificial intelligence (AI) continues to permeate society, there has been an ever-increasing demand for efficient AI hardware, accelerators and integrated circuits (ICs). The field of neuromorphic engineering has been instrumental in advancing AI hardware, with concepts like compute-in-memory, pooling, and attention being integrated into mainstream IC designs. However, to sustain the growth of the field, there is a pressing need to address the education and workforce gap in the discipline of IC design and fabrication. By leveraging the popularity of AI, machine learning and neuromorphic engineering disciplines, the project is motivating researchers to acquire design skills that can then be applied to various domains of IC design. At the centerpiece of this RCN is the Annual Telluride Neuromorphic Cognition Engineering (TNCE) workshop which is an influential and highly anticipated event in the field of neuroscience and artificial intelligence. This workshop brings together leading researchers, experts, and students from various disciplines to explore and advance our understanding of neuromorphic computing and cognition. As a part of this RCN, the project is leveraging the workshop infrastructure to organize discussion groups, facilitate hands-on training events, and form sustainable research cohorts around the theme of neuromorphic ICs.The intellectual merit of the RCN is to understand and exploit emergent neurodynamics and noise-induced processes in neuromorphic devices and hardware to efficiently solve specific computing tasks. To achieve this the RCN is utilizing the extensive expertise within the network to investigate novel neuromorphic architectures, circuits and hardware that could lead to significant performance advantages when compared to other computing architectures that use central processing units (CPUs), graphical processing units (GPUs) or quantum processors. The RCN consists of three coordination sub-networks: (a) The NICE-RCN Co-design Framework is concentrating on the development of behavioral models and software. These tools will enable the simulation and evaluation of different architectures on benchmark tasks; (b) The NICE-RCN IC Design Framework will focus on designing neuromorphic Process Design Kits (PDKs) using open-source IC fabrication tools accessible through Efabless, as well as the Taiwan Semiconductor Manufacturing Corporation (TSMC) PDK tools provided by Muse Semiconductors; (c) The NICE-RCN IC Testing Framework will establish a testing infrastructure and create benchmarks for evaluating the performance of neuromorphic ICs.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.

该项目正在促进神经形态集成电路教育（NICE）研究协调网络（RCN），这将降低对神经形态集成电路（IC）的设计，制造和测试感兴趣的学生和研究人员的进入门槛。随着人工智能（AI）继续渗透到社会中，对高效AI硬件、加速器和集成电路（IC）的需求不断增加。神经形态工程领域在推进人工智能硬件方面发挥了重要作用，内存计算、池化和注意力等概念被集成到主流IC设计中。然而，为了维持该领域的增长，迫切需要解决IC设计和制造学科的教育和劳动力缺口。通过利用人工智能，机器学习和神经形态工程学科的普及，该项目激励研究人员获得设计技能，然后应用于IC设计的各个领域。本次RCN的核心是年度碲化物神经形态认知工程（TNCE）研讨会，这是神经科学和人工智能领域有影响力和备受期待的事件。本次研讨会汇集了来自各个学科的领先研究人员，专家和学生，以探索和推进我们对神经形态计算和认知的理解。作为该RCN的一部分，该项目利用研讨会基础设施组织讨论小组，促进实践培训活动，并围绕神经形态ICs. RCN的智力价值的主题形成可持续的研究队列是理解和利用神经形态设备和硬件中的紧急神经动力学和噪声诱导过程，以有效地解决特定的计算任务。为了实现这一目标，RCN正在利用网络内广泛的专业知识来研究新型神经形态架构、电路和硬件，与使用中央处理器（CPU）、图形处理器（GPU）的其他计算架构相比，这些架构可能会带来显着的性能优势。或量子处理器。区域网络由三个协调子网络组成：（a）NICE-RCN共同设计框架集中于行为模式和软件的开发。（B）NICE-RCN集成电路设计框架将侧重于使用通过Efabless获得的开源集成电路制造工具以及Muse Semiconductors提供的台湾半导体制造公司（台积电）PDK工具设计神经形态工艺设计工具包（PDK）;（c）NICE-RCN集成电路测试框架将建立一个测试基础设施，并为评估神经形态集成电路的性能建立基准。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。