RTML: Small: Achieving Real-Time and Energy-efficient Computing for 5G Networks (ARTEN): A Deep Reservoir Computing Approach

RTML：小型：实现 5G 网络的实时和节能计算 (ARTEN)：一种深水库计算方法

• 批准号: 1937487
• 负责人: Yang Yi
• 金额: $ 50万
• 依托单位: Virginia Polytechnic Institute and State University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-10-01 至 2024-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1937487&HistoricalAwards=false
• 关键词: RTML; Small; Achieving; Real; Time

Fueled by the popularity of smartphones as well as the upcoming deployment of the fifth-generation (5G) mobile broadband wireless networks, mobile data traffic is projected to have an explosive growth in the near future. In 5G wireless networks, enhanced mobile broadband (eMBB) and massive machine type communications (mMTC) are regarded as two primary use cases for 5G radios where eMBB supports stable connections with very high data rates for low-speed mobile users as well as moderate data rates for high-speed mobile users, and eMTC supports a massive number of Internet of Things (IoT) devices which are only sporadically active to transmit small data payloads. In this project, novel machine-learning-based hardware and customized efficient training algorithms tailored towards 5G eMBB and mMTC will be introduced to achieve real-time and energy-efficient computing for 5G wireless networks. The proposed research will involve training of both graduate and undergraduate students in circuit design, computing, communications, and networking. Through the development of application-oriented projects and various outreach activities supported by Virginia Tech's outreach programs, undergraduates including students from underrepresented groups in STEM will have opportunities to develop creative and independent problem-solving skills. The objective of this project is to develop a novel deep-learning-based hardware and software co-design platform to achieve real-time and energy-efficient computing for 5G wireless networks focusing on eMBB and mMTC. To be specific, 1) deep-learning-based integrated circuits and architectures will be designed to achieve real-time and energy-efficient learning; and 2) hardware-software-algorithm co-design with customized low-complexity online training algorithms will be introduced to realize high data rate symbol detection for eMBB as well as to realize distributed contention-based random-access strategies for mMTC. Both software and hardware testbeds will be developed to evaluate proposed hardware designs and learning algorithms. The success of the project will potentially revolutionize the telecommunication industry by incorporating machine learning and deep learning to the future 5G wireless devices and networks.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.

在智能手机的普及以及即将部署的第五代(5G)移动宽带无线网络的推动下，移动数据流量预计在不久的将来将出现爆炸性增长。在5G无线网络中，增强型移动宽带(EMBB)和大规模机器类型通信(MMTC)被视为5G无线电的两个主要使用案例，其中eMBB支持低速移动用户具有非常高数据速率的稳定连接，而高速移动用户支持中等数据速率的稳定连接，eMTC支持大量仅偶尔活跃地传输小数据有效载荷的物联网(IoT)设备。在本项目中，将引入新的基于机器学习的硬件和定制的针对5G eMBB和MMTC的高效训练算法，以实现5G无线网络的实时和节能计算。拟议的研究将涉及电路设计、计算、通信和网络方面的研究生和本科生的培训。通过开发以应用为导向的项目和由弗吉尼亚理工大学外展项目支持的各种外展活动，本科生，包括STEM中代表性不足群体的学生，将有机会发展创造性和独立解决问题的技能。本项目的目标是开发一种新型的基于深度学习的软硬件协同设计平台，以实现以eMBB和MMTC为核心的5G无线网络的实时节能计算。具体地说，1)设计了基于深度学习的集成电路和结构，以实现实时和能量有效的学习；2)硬件-软件-算法协同设计和定制的低复杂度在线训练算法，将实现eMBB的高数据率符号检测以及MMTC的基于竞争的分布式随机接入策略。将开发软件和硬件试验台，以评估拟议的硬件设计和学习算法。该项目的成功将潜在地通过将机器学习和深度学习融入未来的5G无线设备和网络来彻底改变电信行业。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

Spatial-Temporal Hybrid Neural Network With Computing-in-Memory Architecture

• DOI: 10.1109/tcsi.2021.3071956
• 发表时间: 2021-04
• 期刊: IEEE Transactions on Circuits and Systems I: Regular Papers
• 作者: Kangjun Bai;Lingjia Liu;Y. Yi

Enhancing SNN Training Performance: A Mixed-Signal Triplet Reconfigurable STDP Circuit with Multiplexing Encoding

增强 SNN 训练性能：具有复用编码的混合信号三元组可重构 STDP 电路

• DOI: 10.1109/iscas46773.2023.10181729
• 发表时间: 2023
• 期刊: 2023 IEEE International Symposium on Circuits and Systems (ISCAS
• 作者: Zheng, Honghao;Yi, Yang
• 通讯作者: Yi, Yang

Enabling a New Methodology of Neural Coding: Multiplexing Temporal Encoding in Neuromorphic Computing

• DOI: 10.1109/tvlsi.2023.3234514
• 发表时间: 2023-03
• 期刊: IEEE Transactions on Very Large Scale Integration (VLSI) Systems
• 影响因子: 2.8
• 作者: Honghao Zheng;Kangjun Bai;Y. Yi

Robust Deep Reservoir Computing Through Reliable Memristor With Improved Heat Dissipation Capability

• DOI: 10.1109/tcad.2020.3002539
• 发表时间: 2021-03-01
• 期刊: IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS
• 影响因子: 2.9
• 作者: An, Hongyu;Al-Mamun, Mohammad Shah;Yi, Yang
• 通讯作者: Yi, Yang

Memristor-based Deep Spiking Neural Network with a Computing-In-Memory Architecture

• DOI: 10.1109/isqed54688.2022.9806206
• 发表时间: 2022-04
• 期刊: 2022 23rd International Symposium on Quality Electronic Design (ISQED)
• 作者: Fabiha Nowshin;Y. Yi