期刊
IEEE TRANSACTIONS ON ELECTRON DEVICES
卷 69, 期 5, 页码 2353-2359出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/TED.2022.3159508
关键词
Neurons; Synapses; Spintronics; Magnetic tunneling; Neuromorphic engineering; Magnetic domain walls; Mathematical models; Artificial neural network; leaky integrate-and-fire (LIF) neuron; multilayer perceptron; neuromorphic computing
资金
- National Science Foundation under Division of Computing and Communication Foundations (CCF) [1910800, 1910997]
- National Science Foundation Graduate Research Fellowship [1746053]
- Eugene McDermott Graduate Fellowship [202001]
- Texas Analog Center of Excellence Graduate Fellowship
- Direct For Computer & Info Scie & Enginr
- Division of Computing and Communication Foundations [1910800] Funding Source: National Science Foundation
- Division of Computing and Communication Foundations
- Direct For Computer & Info Scie & Enginr [1910997] Funding Source: National Science Foundation
CMOS devices are not suitable for analog applications, while spintronic devices are well suited. However, a large number of spintronic devices still require the use of CMOS, which decreases system efficiency. This study improves the activation functions of spintronic devices to enable better neural network learning and recognition.
CMOS devices display volatile characteristics and are not well suited for analog applications such as neuromorphic computing. Spintronic devices, on the other hand, exhibit both non-volatile and analog features, which are well suited to neuromorphic computing. Consequently, these novel devices are at the forefront of beyond-CMOS artificial intelligence applications. However, a large quantity of these artificial neuromorphic devices still require the use of CMOS to implement various neuromorphic functionalities, which decreases the efficiency of the system. To resolve this, we have previously proposed a number of artificial neurons and synapses that do not require CMOS for operation. Although these devices are a significant improvement over previous renditions, their ability to enable neural network learning and recognition is limited by their intrinsic activation functions. This work proposes modifications to these spintronic neurons that enable configuration of the activation functions through control of the shape of a magnetic domain wall track. Linear and sigmoidal activation functions are demonstrated in this work, which can be extended through a similar approach to enable a wide variety of activation functions.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据