碳纳米管
材料科学
时间常数
瞬态(计算机编程)
碳纳米纤维
纳米纤维
油藏计算
光电子学
MNIST数据库
纳米技术
脉搏(音乐)
电压
纤维素
响应时间
计算机科学
常量(计算机编程)
恒压
上升时间
瞬态响应
碳纤维
纳米电子学
作者
Kotaro Takanashi,Ryo Tagami,Hiroaki Komatsu,Takashi Ikuno
标识
DOI:10.35848/1347-4065/ae5ab2
摘要
We fabricated paper-based synaptic devices composed of multi-walled carbon nanotubes (MWCNTs) and cellulose nanofibers (CNFs), aiming for transient responses on human time scales from sub-seconds to seconds. Their voltage responses to pulsed inputs were investigated for physical reservoir computing (PRC) applications. The devices exhibit transient responses with characteristic time constants ranging from sub-millisecond to several seconds, systematically tunable via the MWCNT concentration. Under consecutive pulse stimulation, they show paired-pulse facilitation, indicating short-term synaptic plasticity. PRC performance was evaluated using short-term memory (STM), parity-check (PC), and handwritten digit recognition tasks. A device with 0.55 wt% MWCNT achieved an STM capacity C STM of 2.03 and a PC capacity C PC of 0.01, and reached 90.2% accuracy in the MNIST recognition task. These results demonstrate that MWCNT/CNF synaptic devices provide a simple, low-cost, and flexible platform for PRC operating on human time scales and are promising for lightweight edge-AI hardware.
科研通智能强力驱动
Strongly Powered by AbleSci AI