权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

Skyrmionic Nanodevices for Neuromorphic Computing

用于神经形态计算的斯格明离子纳米器件

基本信息

批准号：
2480931
负责人：
金额：
--
依托单位：
University of Manchester
依托单位国家：
英国
项目类别：
Studentship
财政年份：
2020
资助国家：
英国
起止时间：
2020 至无数据
项目状态：
已结题

来源：
https://gtr.ukri.org/projects?ref=studentship-2480931
关键词：
Skyrmionic Nanodevices Neuromorphic Computing

项目摘要

The project aims to bridge the burgeoning field of skyrmionics with neuromorphic computing in order to try and bring new solutions from the world spintronics into neuromorphic hardware. In particular, the goal of the project is to develop realistic dense skyrmionic nanodevices that can serve as non-volatile and energy-efficient neuromorphic hardware components that could be deployed in edge-computing scenarios. In order to achieve this, we will use i) micromagnetic simulations to design the devices and exploit the intriguing properties of skyrmionic materials, ii) circuit simulations in order to investigate how to connect these devices in a hybric spintronics-CMOS approach and iii) spiking neural network simulations in order to explore their integration into a neuromorphic computing paradigm. The project aims to design novel neuromorphic components and to evaluate their potential, performance and energy-efficient operation if embedded in full systems/chips for edge-computing applications.Research questions / Objectives:- Investigate skyrmionic device concepts that emulate synapses or neurons with smaller footprint and higher energy-efficiency.- Investigate skyrmionic interconnects.- Investigate compact circuit models that interface skyrmionic nanodevices with CMOS circuits.- Investigate optimal spiking neural network topologies for integrating skyrmionics hardware components.Approach / Methodologies:- Computational nanophysics simulations, micromagnetics and atomistic simulations- CMOS circuit simulations- Spiking Neural Network SimulationsNovel physical sciences/engineering content:The project is interdisciplinary and combines research threads from various EPSRC research areas: - spintronics- artificial intelligence technologies- Condensed matter: magnetism and magnetic materials- Pervasive and ubiquitous computing

该项目旨在将新兴的天电子学领域与神经形态计算联系起来，以尝试将世界自旋电子学的新解决方案引入神经形态硬件。特别是，该项目的目标是开发逼真的高密度天空微子纳米设备，这些设备可以作为非易失性和高能效的神经形态硬件组件，可以部署在边缘计算场景中。为了实现这一目标，我们将使用i)微磁模拟来设计器件并探索天空微电子材料的有趣特性，ii)电路模拟以研究如何以混合自旋电子学-CMOS方法来连接这些器件，以及iii)尖峰神经网络模拟以探索它们与神经形态计算范式的集成。该项目旨在设计新的神经形态组件，并评估如果将其嵌入到用于边缘计算的完整系统/芯片中，它们的潜力、性能和节能操作。研究问题/目标：-研究以更小的占地面积和更高的能量效率模拟突触或神经元的天光设备概念。-研究天光设备互连。-研究将天光纳米设备与CMOS电路连接的紧凑电路模型。-研究集成天光电子硬件组件的最佳尖峰神经网络拓扑。方法/方法：-计算纳米物理模拟，微磁学和原子模拟-cmos电路模拟-尖峰神经网络模拟新颖的物理科学/工程学内容：该项目是跨学科的，结合了EPSRC各个研究领域的研究思路：-自旋电子学-人工智能技术-凝聚态：磁性和磁性材料-普适和普适计算