基于非易失性存储器件的类脑计算研究进展
Research progress of nuero-inspired computing using non-volatile memory
-
摘要: 随着人工智能领域的快速发展,海量数据的处理需要相应硬件的对应开发。对人脑物理结构及工作机制的深入研究使得基于非易失性内存的类脑计算器件集成一系列受神经系统启发的功能,从而为大数据计算工作提供一种高效、节能的方法。本文介绍了类脑计算的发展历程,概述了针对不同应用的非易失性内存器件的大致分类,并着重介绍了代表性的非易失性内存器件构建人工突触和神经元器件的研究进展。Abstract: The rapid development of artificial intelligence demands the development of relevant hardware processing massive quantity of data. Neuro-inspired computing chips integrate a range of features inspired by neuro systems and could provide an energy-efficient approach to computing workloads involving big data. This article reviews the development history of neuro-inspired computing and generally introduces the genres of the non-volatile memory(NVM) devices based on their various applications. Then the article further reviewed the state of the art of the neuromorphic researches in which the NVM devices play the role of either the synapse or the neuron.