CAREER: Scalable Ionic Gated 2D Synapse (IG-2DS) with Programmable Spatio-Temporal Dynamics for Spiking Neural Networks

职业：具有可编程时空动力学的可扩展离子门控 2D 突触 (IG-2DS)，适用于尖峰神经网络

• 批准号: 1943683
• 负责人: Feng Xiong
• 金额: $ 50万
• 依托单位: University of Pittsburgh
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2020
• 资助国家: 美国
• 起止时间: 2020-03-01 至 2025-02-28
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1943683&HistoricalAwards=false
• 关键词: CAREER; Scalable; Ionic; Gated; 2D

CAREER: Scalable Ionic Gated 2D Synapse (IG-2DS) with Programmable Spatio-Temporal Dynamics for Spiking Neural NetworksECCS : 1943683PI: Feng XiongNontechnical:Artificial intelligence (AI) has the potential to transform people's lives. Machine learning has shown promise in healthcare, transportation, and advanced manufacturing, but requires enormous amounts of energy. The human brain is better at cognitive tasks such as pattern recognition than even supercomputers while requiring less than 20W of power. Inspired by the human brain, neuromorphic computing and artificial neural networks have recently attracted immense interest. Spiking neural networks mimic biological processes by incorporating similar temporal dynamics. This type of computing offers a promising route for energy-efficient computing with high bandwidth. It is, however, challenging and expensive to implement spatio-temporal processes such as short-term and long-term memory in spiking neural networks with existing digital electronics. In this project, the PI will develop a critically missing element, a dynamic synapse, by controlling the charge carrier concentration in two-dimensional devices. This breakthrough will lead to a truly neuro-realistic computing system with dramatic improvements in energy efficiency, bandwidth, and cognitive capabilities. This can lead to the wide use of AI and revolutionize society through advances in cognitive computing, self-driving vehicles, and autonomous manufacturing. The PI will develop an afterschool outreach program in partnership with a local community engagement center. Students from underrepresented groups will have design and laboratory experiences with the aim of attracting them into engineering careers.Technical:The objective of this project is to elucidate the transport mechanisms in ionic gated two-dimensional synaptic (IG-2DS) devices and build synaptic arrays with programmable spatio-temporal dynamics, high precision, low power, good scalability, and good reliability for the hardware implementation of spiking neural networks (SNNs). Despite recent success in the development of artificial neural networks, the hardware implementation of SNNs has been challenging because existing digital electronics do not possess the spatio-temporal dynamics needed for a dynamic synapse-the key building block of SNNs. In this project, the PI will adopt a novel approach to demonstrate short-term and long-term plasticity in 2D synapses by modulating the volatile doping effect from ionic gating and the non-volatile doping effect from charge transfer doping via intercalation. The PI will carry out the following three research tasks: (1) elucidate the short- and long-term doping mechanisms in IG-2DS; (2) demonstrate tunable spatio-temporal dynamics in IG-2DS for the hardware implementation of SNNs; and (3) investigate the scaling potentials of IG-2DS for large-scale integration. Fundamentally, this work elucidates the electronic and ionic transport in 2D electronics. Practically, this work will have an immense impact in the fields of computing, self-driving, automation in manufacturing, flexible sensors, and wearable electronics through the hardware implementation of SNNs.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.

职业：可扩展的离子门控2D突触（IG-2DS）与可编程的时空动态尖峰神经网络ECCS：1943683 PI：冯雄非技术：人工智能（AI）有可能改变人们的生活。机器学习在医疗保健、交通运输和先进制造业中显示出了希望，但需要大量的能源。人类大脑在模式识别等认知任务方面甚至比超级计算机更好，而所需功率不到20瓦。受人脑的启发，神经形态计算和人工神经网络最近引起了人们极大的兴趣。脉冲神经网络通过结合类似的时间动态来模拟生物过程。这种类型的计算为高带宽的节能计算提供了一条有前途的路线。然而，这是具有挑战性的和昂贵的实现时空过程，如短期和长期记忆的尖峰神经网络与现有的数字电子。在这个项目中，PI将通过控制二维设备中的电荷载流子浓度来开发一个关键的缺失元素，一个动态突触。这一突破将导致一个真正的神经现实的计算系统，在能源效率，带宽和认知能力的显着改善。这可能导致人工智能的广泛使用，并通过认知计算、自动驾驶汽车和自动制造的进步来彻底改变社会。PI将与当地社区参与中心合作制定课后外展计划。技术：本项目的目标是阐明离子门控二维突触（IG-2DS）器件中的传输机制，并构建具有可编程时空动力学、高精度、低功耗、良好可扩展性和良好可靠性的突触阵列，用于脉冲神经网络（SNN）的硬件实现。尽管最近在人工神经网络的发展取得了成功，SNN的硬件实现一直具有挑战性，因为现有的数字电子不具备动态突触所需的时空动态SNN的关键构建块。在这个项目中，PI将采用一种新的方法，通过调节离子门控的挥发性掺杂效应和通过嵌入的电荷转移掺杂的非挥发性掺杂效应，来证明2D突触的短期和长期可塑性。主要研究员将进行以下三项研究工作：（1）阐明IG-2DS的短期和长期掺杂机制;（2）展示IG-2DS中用于SNN硬件实现的可调时空动力学;以及（3）研究IG-2DS用于大规模集成的缩放潜力。从根本上说，这项工作阐明了电子和离子的二维电子输运。实际上，通过SNN的硬件实现，这项工作将在计算、自动驾驶、制造自动化、柔性传感器和可穿戴电子产品领域产生巨大影响。该奖项反映了NSF的法定使命，通过使用基金会的智力价值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

Building Time-Surfaces by Exploiting the Complex Volatility of an ECRAM Memristor

• DOI: 10.1109/jetcas.2023.3330832
• 发表时间: 2022-01
• 期刊: IEEE Journal on Emerging and Selected Topics in Circuits and Systems
• 影响因子: 4.6
• 作者: Marco Rasetto;Qingzhou Wan;Himanshu Akolkar;Bertram E. Shi;Feng Xiong;R. Benosman