In recent years, neuromorphic computing has been intensively investigated, to take over the conventional or von Neumann scheme. Herein, the advantages of memristors as neurons and synapses are discussed. After a brief introduction to biological neurons and synapses, focus is put on spin-based devices, including magnetic tunnel junction (MTJ) and domain wall devices. Certain materials and device designs aim at mimicking synapses’ functionality, whereas others gather both neurons and synapses. The advancements in spin-based memory applications are of great advantage for neuromorphic computing and their implementation is presented herein.

中文翻译：

---

用于神经形态计算的多态磁畴壁设备

近年来，神经拟态计算得到了深入研究，以取代传统或冯诺依曼方案。在此，讨论了忆阻器作为神经元和突触的优势。在简要介绍了生物神经元和突触之后，重点介绍了基于自旋的器件，包括磁隧道结 (MTJ) 和畴壁器件。某些材料和设备设计旨在模仿突触的功能，而另一些则同时收集神经元和突触。基于自旋的存储器应用的进步对神经形态计算具有很大的优势，本文介绍了它们的实现。