%0 Journal Article
%T 基于ZrO<sub>2</sub>/ZnO异质结人工突触的研究及应用<br>Research and Application of Artificial Synapses Based on ZrO<sub>2</sub>/ZnO Heterojunction
%A 任齐重
%J Artificial Intelligence and Robotics Research
%P 510-518
%@ 2326-3423
%D 2025
%I Hans Publishing
%R 10.12677/airr.2025.143050
%X 由忆阻器构建的类脑神经形态网络凭借其低功耗、高效的计算机架构而引起了广泛的关注&#65292;也有望打破传统的冯&#183;诺伊曼设计的局限性。本文采用射频磁控溅射技术制备了一种基于ZrO<sub>2</sub>/ZnO异质结的人工突触器件&#65292;该器件可以实现典型的突触行为&#65292;包括短期可塑性、长期可塑性和成对脉冲促进行为。此外&#65292;该器件构建的卷积神经网络对手写数字数据集和时尚服装数据集分别具有96.6%和71.1%的识别率。这些结果表明&#65292;该设备在神经形态计算领域具有巨大的潜力&#65292;为今后建立有效的神经形态计算系统提供了一种可行的方法。<br>The brain-like neuromorphic network constructed by memristors has attracted wide attention due to its low power consumption and efficient computer architecture, and is also expected to break the limitations of traditional von Neumann design. In this paper, an artificial synaptic device based on ZrO<sub>2</sub>/ZnO heterojunction was prepared by RF magnetron sputtering technology. The device can achieve typical synaptic behavior, including short-term plasticity, long-term plasticity and paired pulse promotion behavior. In addition, the convolutional neural network constructed by the device has a recognition rate of 96.6% and 71.1% for handwritten digital data sets and fashion clothing data sets, respectively. These results show that the device has great potential in the field of neuromorphic computing and provides a feasible method for establishing an effective neuromorphic computing system in the future.
%K 忆阻器&#65292
%K 人工突触器件&#65292
%K 神经形态计算<br>Memristor
%K Artificial Synaptic Device
%K Neuromorphic Computing
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=113940