E2CDA: Type I: Collaborative Research: Nanophotonic Neuromorphic Computing

E2CDA：I 型：协作研究：纳米光子神经形态计算

• 批准号: 1740235
• 负责人: Volker Sorger
• 金额: $ 55.29万
• 依托单位: George Washington University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-09-15 至 2021-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1740235&HistoricalAwards=false
• 关键词: E2CDA; Type; Collaborative; Research; Nanophotonic

With this program, the National Science Foundation has challenged the scientific community to help solve the problem of energy efficient computing. Data centers around the world spend almost equal amount of overhead power in cooling and power conversion for every Watt of computation. As society's appetite for information continues to grow, data centers will eventually consume a significant portion of the world's electricity. Today, the entire information technology industry relies on digital electronics, i.e. electrons carrying digital bits, which impose a barrier for how efficient a computational task can be, regardless of how brilliantly software routines can be engineered. This project will explore the physics of lightwaves to craft a "photonic" processor that breaks that efficiency barrier. This interdisciplinary endeavor will attempt to generate foundational science which will support future innovation in the tech industry. To the end of help building the workforce for the future, the investigators will continue their commitments in mentoring graduate and undergraduate students, and support broadening participation of women, minorities and underrepresented groups through STEM outreach, and NSF REU sites.The approach unites neuromorphic (neural network-inspired) architectures, nanophotonic devices, and emerging fabrication platforms. The technical challenge requires vertical integration of new optical devices, logic units, and control software. It utilizes diverse expertise of faculty members in nanophotonic devices, photonic systems and computer engineering. Specifically, the key points of innovation will include: 1) ultra-efficient electro-optic modulators that will act as neurons at the nanoscale; 2) reconfigurable hardware interconnection fabric on-chip that uses light as information carrier; and 3) creation of control software and benchmarks that are required to orient the project within the broader field of efficient computing. A strong experimental thrust to design, build, and demonstrate these proposed systems will serve to reduce the risk of development of this technology and pave the way for other researchers to join the emerging field of nanophotonic neuromorphic computing.

通过这个项目，国家科学基金会向科学界提出了挑战，以帮助解决节能计算的问题。世界各地的数据中心在每瓦特计算的冷却和功率转换方面花费几乎相同的开销功率。随着社会对信息的需求不断增长，数据中心最终将消耗世界电力的很大一部分。今天，整个信息技术行业都依赖于数字电子，即携带数字比特的电子，这对计算任务的效率造成了障碍，无论软件例程的设计多么出色。该项目将探索光波的物理学，以制造一种“光子”处理器，打破效率障碍。这种跨学科的奋进将试图产生基础科学，这将支持未来的创新技术产业。为了帮助构建未来的劳动力，研究人员将继续致力于指导研究生和本科生，并通过STEM推广和NSF REU网站支持扩大女性，少数民族和代表性不足的群体的参与。该方法将神经形态（神经网络启发）架构，纳米光子设备和新兴制造平台结合起来。技术挑战需要新的光学器件、逻辑单元和控制软件的垂直集成。它利用教师在纳米光子器件，光子系统和计算机工程方面的各种专业知识。具体而言，创新的关键点将包括：1）超高效电光调制器，将在纳米级充当神经元; 2）可重新配置的芯片上硬件互连结构，使用光作为信息载体;以及3）创建控制软件和基准，这些软件和基准是在更广泛的高效计算领域内定位项目所需的。一个强大的实验推力来设计，构建和演示这些拟议的系统将有助于降低这种技术的发展风险，并为其他研究人员加入新兴的纳米光子神经形态计算领域铺平道路。

项目成果

Roadmap on material-function mapping for photonic-electronic hybrid neural networks

• DOI: 10.1063/1.5109689
• 发表时间: 2019-10-01
• 期刊: APL MATERIALS
• 影响因子: 6.1
• 作者: Miscuglio, Mario;Adam, Gina C.;Sorger, Volker J.
• 通讯作者: Sorger, Volker J.

Advances in photonic neuromorphic computing (Conference Presentation)

光子神经形态计算的进展（会议演讲）

• DOI: 10.1117/12.2509838
• 发表时间: 2019
• 期刊: Smart Photonic and Optoelectronic Integrated Circuits XXI
• 作者: Sorger, Volker J.;George, Jonathan K.;Mehrabian, Armin;Shastri, Bhavin;El-Ghazawi, Tarek;Prucnal, Paul R.;Lee, El-Hang;He, Sailing
• 通讯作者: He, Sailing

Energy-efficient graphene and ITO-based MZI and absorption modulators (Conference Presentation)

节能石墨烯和基于 ITO 的 MZI 和吸收调制器（会议演示）

• DOI: 10.1117/12.2509723
• 发表时间: 2019
• 期刊: Optical Components and Materials XVI
• 作者: Sorger, Volker J.;Amin, Rubab;Ma, Zhizhen;Chen, Ray;Dalir, Hamed;Digonnet, Michel J.;Jiang, Shibin
• 通讯作者: Jiang, Shibin

Neuromorphic photonics with electro-absorption modulators

• DOI: 10.1364/oe.27.005181
• 发表时间: 2019-02-18
• 期刊: OPTICS EXPRESS
• 影响因子: 3.8
• 作者: George, Jonathan K.;Mehrabian, Armin;Sorger, Volker J.
• 通讯作者: Sorger, Volker J.

A Design Methodology for Post-Moore’s Law Accelerators: The Case of a Photonic Neuromorphic Processor

后摩尔定律加速器的设计方法：光子神经形态处理器的案例

• DOI: 10.1109/asap49362.2020.00028
• 发表时间: 2020
• 期刊: Architectures and Processors (ASAP
• 作者: Mehrabian, Armin;Sorger, Volker J;El-Ghazawi, Tarek
• 通讯作者: El-Ghazawi, Tarek