Neuromorphic computing with antiferromagnetic spintronics
Neuromorphic computing with antiferromagnetic spintronics
复制标题
反铁磁自旋电子学神经形态计算
DOI：
10.1063/5.0009482
复制
发表时间：
2020-07-07
影响因子：
3.2
通讯作者：
Ohno, Hideo
中科院分区：
文献类型：
--
作者：
Kurenkov, Aleksandr;Fukami, Shunsuke;Ohno, Hideo
虽然能够轻松解决认知任务的人工智能已经改变了技术和日常生活，但在单位能量性能方面与生物系统仍然存在巨大差距。神经形态计算，其中硬件与替代架构，电路，设备，和/或材料的探索，预计将减少差距。反铁磁自旋电子学可以为这一方案提供一个有前途的平台。反铁磁系统的主动功能最近已被证明，一些工作表明其潜在的生物启发计算。在这个角度来看，我们期待通过这些作品的棱镜和讨论的前景和挑战的反铁磁自旋电子学的神经形态计算。综述和讨论了非脉冲人工神经网络，脉冲神经网络和水库计算。
While artificial intelligence, capable of readily addressing cognitive tasks, has transformed technologies and daily lives, there remains a huge gap with biological systems in terms of performance per energy unit. Neuromorphic computing, in which hardware with alternative architectures, circuits, devices, and/or materials is explored, is expected to reduce the gap. Antiferromagnetic spintronics could offer a promising platform for this scheme. Active functionalities of antiferromagnetic systems have been demonstrated recently and several works indicated their potential for biologically inspired computing. In this perspective, we look through the prism of these works and discuss prospects and challenges of antiferromagnetic spintronics for neuromorphic computing. Overview and discussion are given on non-spiking artificial neural networks, spiking neural networks, and reservoir computing.