脉冲神经网络模式收敛与学习初步研究

论文总字数：32793字

摘 要

随着人工智能、机器学习的深入发展，对神经网络的研究也不断取得进展。近些年来，新一代的神经网络模型:脉冲神经网络开始得到人们越来越多的关注。这不仅仅因为其相对于传统神经网络所拥有的许多优势，还因为这种神经网络的复杂性，随着解剖学、生理学等的发展不断有新的基于生物学模型的理论提出。

本文主要内容有四点。第一章绪论简单交代了课题的背景，包括神经网络的发展历史、传统神经网络的基本模型以及脉冲神经网络的简介。第二章主要内容是有关神经网络生物学模型所具备的一些特征，从其结构、功能以及部分与信息处理有关的内容都有涉及。第三章对脉冲神经网络神经元的特性作了简单介绍，并将现有的一些常用的神经元模型进行了比较。第四章讲述了在matlab上对脉冲神经网络进行仿真的过程，包括仿真单个神经元、仿真整个神经网络、加入STDP学习规则的详细步骤，并对不同输入下的仿真结果进行了简单的收敛性分析，实现了简单的记忆功能。在论文的最后，本文讨论了此次毕业设计的意义以及在此基础上可以继续进行的工作。

关键词：脉冲神经网络，神经元模型，学习规则，神经网络仿真,STDP

STUDY IN PATTERN CONVERGENCY AND PATTERN LAERNING OF THE SPIKING NEURAL NETWORK

Abstract

With the development of artificial intelligence and machine learning, the research on neural network research is also progressing. In recent years, a new generation of neural network model: Spiking Neural Network model, has begun to gain more and more attention. This is not only because of its many advantages relative to the traditional neural networks, but also because of the complexity of this neural network. As disciplines such as anatomy, physiology, etc. develops, many new biological-based SNN models comes out.

There are four main contents in this article. The first chapter briefly talked about the background of the subject, including a brief description of the history of the development of neural networks, the model of the traditional neural network and a introduction to spiking neural network. The second chapter is about the features the biological neural network model has, specifying in its structure, functions and some related content on information processing. The third chapter of the characteristics of pulse neural network neurons made a brief introduction, and some of the existing common neuron models were compared. The fourth chapter describes the procedure of simulating SNN network using matlab, including the simulation of individual neurons, the simulation of the entire neural network simulation, applying STDP learn rules to the network, and a simple analysis of the final result from different input. A simple implementation of memory function was also done in this chapter. In the end of the paper, there are some discussions about the significance of this particular graduation design work, and some advices were made on what further work could be done.

KEY WORDS: Pulse neural network, izhikevich, STDP, matlab

目 录

摘 要 I

Abstract II

第一章 绪 论 1

1.1 脉冲神经网络的研究背景及意义 1

1.1.1 研究背景 1

1.1.2 三代神经网络简介 1

1.2 人工神经网络简介 2

1.2.1 人工神经网络神经元模型 2

1.3 脉冲神经网络简介 2

第二章 神经突触生理学模型 4

第三章 脉冲神经网络模型以及学习规则的选取 6

3.1 计算神经网络的特点 6

3.2 常用的脉冲神经网络模型 8

3.2 STDP学习机制 11

第四章 脉冲神经网络仿真及分析 13

4.1 Izhikevich模型神经元的仿真 13

4.1.1 以外部电流作为输入的izhikevich神经元仿真 13

4.1.2 以突触电流作为输入的izhikevich神经元仿真 14

4.2 双层izhikevich神经网络的仿真 15

4.3 加入STDP学习规则后的双层izhikevich神经网络仿真 16

4.4 加入STDP学习规则后仿真结果分析 17

第五章 总结与展望 21

参考文献 22

企业教师毕业设计指导工作记录 24

企业教师毕业设计指导评价意见表 27

第一章 绪 论

1.1 脉冲神经网络的研究背景及意义

1.1.1 研究背景

剩余内容已隐藏，请支付后下载全文，论文总字数：32793字