Abstract:

　　Great progress has been achieved in the software implementation of artificial intelligence recently, where "deep learning" is a representative example. The hardware devices for the brain-inspired computing, on the other hand, is just an emergent field of research. The magnetic nanostructures that have been extensively studied in spintronics, such as magnetic tunnel junctions, magnetic domain walls, etc. possess the required physical properties of the elements for brain-inspired computing and are therefore naturally suitable to be used for the hardware implementation of artificial neural networks.

　　We focus on the physical implementation and examination of the brain-inspired computing based upon spintronic devices using micromagnetic simulation combined with the first-principles spin transport calculation. The latter can provide some key parameters for the corresponding magnetic nanostructures. In this talk, I will briefly introduce two examples, i.e. realization of the short-term synaptic plasticity using magnetic tunnel junctions and periodic rhythmic patterns generated by a spintronics recurrent neural network. These studies demonstrate that the artificial neural networks made of spintronic devices can process dynamical information, beyond the focus of the present researches of machine learning–the recognition and classification of static objects.

报告人简介：

　　袁喆，在清华大学物理系就读本科和研究生；获瑞典Chalmers理工大学博士学位；随后在中科院物理研究所，荷兰Twente大学，德国Mainz大学等机构工作；2015年入职北京师范大学；主要从事量子自旋输运、自旋类脑计算的理论与数值计算方面研究。