神经形态工程学
实现(概率)
记忆电阻器
材料科学
MNIST数据库
电导
锂(药物)
人工神经网络
重置(财务)
纳米技术
光电子学
计算机科学
电子工程
人工智能
工程类
数学
物理
内分泌学
经济
金融经济学
统计
医学
凝聚态物理
作者
Anhe Bamao,Yaxin Xia,Guokun Ma,Xiaoxu Yuan,Zisheng Yang,Ao Chen,Chun-Chu Lin,Ting‐Chang Chang,Hao Wang
标识
DOI:10.1016/j.ceramint.2022.11.190
摘要
Devices suitable for neuromorphic computing were urgently needed. Here, by doping lithium metal atoms in solid electrolyte NbOx, a new nonvolatile memory device Pt/LiNbOx/TiN was fabricated and demonstrated typical bipolar resistive storage characteristics. Besides, continuous conductance states were obtained in both DC and pulsed modes and the paired-pulse facilitation (PPF) and spike-timing-dependent plasticity learning (STDP) behavior were successfully simulated by using appropriate pulse sequences, which were derived from the combined action of lithium and oxygen ions. Furthermore, the convolutional neural network(CNN)based on the LiNbOx memristor achieved high accuracy in MNIST dataset, indicating the great potential of the device for applying in neuromorphic computations.
科研通智能强力驱动
Strongly Powered by AbleSci AI