忆阻器的存储特性及其在神经网络方面的应用研究

发布时间：2018-02-15 06:11

本文关键词： 忆阻器人工神经网络可编程神经元霍普菲尔德　出处：《电子科技大学》2015年硕士论文　论文类型：学位论文

【摘要】：近年来,忆阻器因其具有的电阻记忆存储功能、阈值特征、低功耗、结构简单等一系列优点吸引了全世界众多科研人员的关注。典型的忆阻器结构是一种三明治结构:上下两层是导电金属层,充当与外界相连的两个电极。中间的是绝缘层,一般是过渡金属氧化物材料。典型的忆阻器在工作的时候有三个过程:电形成过程(FORMING),置位过程(SET),复位过程(RESET)。所谓FORMING指的是:一个忆阻器件要想实现高低阻态切换的功能,必须在这之前对其进行一次激发阻变材料的操作,这个激发操作之后,忆阻器件才具有正常的忆阻特性;所谓SET指的是:忆阻器由高阻态(HRS)转变到低阻态(LRS)的过程;所谓RESET指的是:忆阻器由低阻态(LRS)转变到高阻态(HRS)的过程。通过对所加的SET电压与RESET电压极性相同与否来划分,忆阻器可分为单极性忆阻器与双极性忆阻器。如果SET与RESET电压同极性,这就是单极性忆阻器,如果不同极性,这就是双极性忆阻器。忆阻器的阻值是可以通过外界的编程电压或者编程电流来进行编程改变的。正是由于这些显著特征,使得忆阻器被广泛应用到存储器、数字电路、混沌电路以及人工神经网络方面。本论文主要对忆阻器在人工神经网络领域的应用进行了深入探讨,提出了用于模拟生物神经元的可编程神经元电路以及在此基础之上设计出来的离散霍普菲尔德神经网络电路。本论文对提出的两种基于忆阻器的新型神经电路先是进行了工作原理的分析,然后对设计的电路结构进行了详细的介绍,并利用Cadence仿真软件对设计的结构进行了仿真,验证结构功能是否正确。最后通过板级电路搭建出离散霍普菲尔德网络电路,并结合FPGA等辅助工具进行了板级电路的测试工作,比对了仿真结果与测试结果,验证了实际设计过程中的可行性。无论是Cadence仿真结果,还是搭建的板级电路测试结果都很明显的反应出:设计的电路功能是正确的,具有集成实现的可行性,对于硬件实现人工神经网络有很好的指导意义。
[Abstract]:In recent years, because of its resistive memory storage function, threshold features, low power consumption, A number of advantages, such as simplicity of structure, have attracted the attention of many researchers around the world. The typical memory device structure is a sandwich structure: the upper and lower layers are conductive metal layers, which act as two electrodes connected to the outside world. It is usually a transition metal oxide material. A typical memory device has three processes at work: the electrical forming process, the setting process, the reset process, and the reset process. The so-called FORMING refers to the function of a memory device to achieve high and low resistive state switching. It is necessary to perform an operation on the material to excite the resistive material before which the device has the normal characteristic of amnesia. The so-called "SET" refers to the process of the transition from the high resistance state to the low resistance state (LRS). The so-called RESET refers to the transition of the resistive device from a low resistance state to a high resistance state. The polarity of the applied SET voltage and the RESET voltage is divided. If SET and RESET voltage are the same polarity, this is unipolar resistor, if different polarity, This is the bipolar resistor. The resistive value of the resistor can be changed by the external programming voltage or the programming current. It is because of these remarkable characteristics that the resistor is widely used in memory, digital circuit, Chaotic circuits and artificial neural networks. This paper mainly discusses the application of amnesia in the field of artificial neural networks. A programmable neuron circuit for simulating biological neurons and a discrete Hopfield neural network circuit based on it are proposed. In this paper, two novel neural circuits based on amnesia are proposed. Is to carry out an analysis of the working principle, Then the structure of the designed circuit is introduced in detail, and the structure is simulated by Cadence software to verify whether the function of the structure is correct. Finally, the discrete Hopfield network circuit is built through the board circuit. Combined with FPGA and other auxiliary tools, the test work of the board level circuit is carried out, and the simulation results and the test results are compared, and the feasibility of the actual design process is verified, regardless of the Cadence simulation results, The test results show that the designed circuit function is correct and has the feasibility of integrated implementation. It has a good guiding significance for the hardware implementation of artificial neural network.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TM501;TP183

【相似文献】