神经形态工程学
冯·诺依曼建筑
材料科学
计算机科学
人工神经网络
瓶颈
记忆电阻器
纳米技术
兴奋剂
各向同性
人工智能
光电子学
电子工程
物理
工程类
嵌入式系统
量子力学
操作系统
作者
Lei Liu,Peng Gao,Mengru Zhang,Juan Dou,Chunsen Liu,Tuo Shi,Hao Huang,Chunlan Wang,Han He,Zijun Chen,Yang Chai,Jianlu Wang,Xuming Zou,Lei Liao,Jingli Wang,Peng Zhou
标识
DOI:10.1002/advs.202408210
摘要
Abstract Neuromorphic computing, a promising solution to the von Neumann bottleneck, is paving the way for the development of next‐generation computing and sensing systems. Axon‐multisynapse systems enable the execution of sophisticated tasks, making them not only desirable but essential for future applications in this field. Anisotropic materials, which have different properties in different directions, are being used to create artificial synapses that can mimic the functions of biological axon‐multisynapse systems. However, the restricted variety and unadjustable conductive ratio limit their applications. Here, it is shown that anisotropic artificial synapses can be achieved on isotropic materials with externally localized doping via electron beam irradiation (EBI) and purposefully induced trap sites. By employing the synapses along different directions, artificial neural networks (ANNs) are constructed to accomplish variable neuromorphic tasks with optimized performance. The localized doping method expands the axon‐multisynapse device family, illustrating that this approach has tremendous potentials in next‐generation computing and sensing systems.
科研通智能强力驱动
Strongly Powered by AbleSci AI