忆阻器类脑神经突触的研究进展

大脑之所以能够控制人和动物的复杂生命活动，使生物体在多变的自然环境得以生存，得益于大规模神经网络中高效、快速、精准的信息传递。神经突触作为神经元之间信息传递的重要机构，保证了神经网络的高效运转，因此构建具有神经突触功能的电子突触是研究仿生系统和类脑神经网络的必经之路。研究人员尝试各种电子元件对神经突触进行模拟，其中忆阻器由于其独特的器件结构和具有“记忆特性”的电学性能，成为构建类脑神经突触的最佳选择。文章全面概述近年来忆阻器模拟神经突触的研究进展，包括忆阻器模拟神经突触的可塑性、再可塑性、非联想学习、联想学习等功能，总结了忆阻器神经突触在人工神经网络中的应用、存在的问题和挑战，并对忆阻器神经突触的研究进行展望。

Research Progress of Memristor-based Neuromorphic Synapses

The efficient, fast and accurate information transmission in the large-scale neural network in the brain is the exact origin that the brain can control the complex life activities of humans and animals and enable organisms to survive in the changeable natural environment. As an important medium for information transmission between neurons, the synapses ensure the efficient operation of neural networks. Therefore, to build electronic synapses with synaptic functions is one essential way to study bionic systems and brain-like neural networks. Researchers have previously tried to simulate synaptic functions with various electronic devices, among which memristor has become one good candidate to build neuromorphic synapses due to its unique device structure and memory characteristics. The researches of memristor-based synapses in recent years are comprehensively summarized in this article, including the synaptic plasticity, metaplasticity, non-associative learning, associative learning and other functions. It also summarizes the application, problems and challenges in artificial neural networks, as well as the research prospects of memristor-based synapses.