神经形态工程学
记忆电阻器
钙钛矿(结构)
计算机科学
MNIST数据库
材料科学
电子工程
纳米技术
钥匙(锁)
电压
人工智能
光电子学
电气工程
计算机体系结构
图层(电子)
现场可编程门阵列
卷积(计算机科学)
机制(生物学)
功率消耗
异质结
领域(数学)
人工神经网络
工程类
可重构性
机器人学
深度学习
嵌入式系统
纳米电子学
电阻随机存取存储器
功率(物理)
实现(概率)
逻辑门
又称作
执行机构
作者
Mingyang Huang,Zhenwang Luo,Hai Wang,Ran Zhao,Xu Wang,Fei Zheng,Zhenfu Zhao,Ziyang Hu
标识
DOI:10.1021/acs.jpclett.5c04053
摘要
s), and endurance (>600 cycles) under ambient conditions. This performance stems from a synergistic "bandgap staircase and built-in electric field" mechanism at the heterointerface, which enhances field confinement for low-voltage switching while the 2D layer suppresses ion migration. Remarkably, this single platform integrates a complete neuromorphic-sensory functional chain. It emulates synaptic plasticity, achieving 90.77% accuracy in MNIST handwritten digit recognition. It also functions as a physical reservoir for temporal pattern processing, reaching perfect classification accuracy. Furthermore, it serves as an artificial nociceptor that faithfully replicates key pain perceptions such as threshold, nonadaptation, sensitization, and relaxation. With ultralow power consumption of only 36 pJ per switch, this multifunctional memristor provides a versatile hardware prototype for next-generation intelligent systems and adaptive human-machine interfaces.
科研通智能强力驱动
Strongly Powered by AbleSci AI