libin毕业论文基于MATLAB神经网络仿真

安徽工业大学管理科学与工程学院基于MATLAB神经网络仿真共45页第1页┊┊┊┊┊┊┊┊┊┊┊┊┊装┊┊┊┊┊订┊┊┊┊┊线┊┊┊┊┊┊┊┊┊┊┊┊┊摘要随着人工神经网络的研究和应用越来越广泛，误差反向传播算法（BP算法）的提出，成功地解决了求解非线性连续函数的多层前馈神经网络权值调整问题，BP神经网络如今成为最广泛使用的网络，研究它对探索非线性复杂问题具有重要意义，而且它具有广泛的应用前景。以BP神经网络为例，讨论了BP神经网络及几种改进BP神经网络性能的算法；通过BP学习算法的推导和分析得知BP网络是一种多层前馈网络，采用最小均方差的学习方式，缺点是仅为有导师训练，训练时间长，易限于局部极小；运用MATLAB来实现各种BP神经网络的实现的设计与训练，比较不同BP神经网络的性能，验证改进BP网络的优势，得出如何根据对象选取神经网络的结论。关键词：人工神经网络、BP神经网络、误差反向传播算法、MATLAB、仿真安徽工业大学管理科学与工程学院基于MATLAB神经网络仿真共45页第2页┊┊┊┊┊┊┊┊┊┊┊┊┊装┊┊┊┊┊订┊┊┊┊┊线┊┊┊┊┊┊┊┊┊┊┊┊┊AbstractWiththeartificialneuralnetworkofresearchandapplicationofmoreandmorewidely,theerrorback-propagationalgorithm(BPalgorithm)isproposed,successfullyresolvedthecontinuousfunctionforsolvingnonlinearmulti-layerfeed-forwardneuralnetworkweightsadjustment,BPnetworkhasbecomenowthemostwidelyusednetworks,Studytoexploreitscomplicatednonlinearproblemhasimportantsignificance,butalsohasbroadapplicationprospects.BPneuralnetworkisdiscussedandseveralimprovementsintheperformanceofBPneuralnetworkalgorithm.BPlearningalgorithmthroughthederivationandanalysisthattheBPnetworkisamulti-layerfeedforwardnetworks,theuseofleast-mean-varianceapproachtolearning,thereisonlydisadvantageisthatthetraininginstructors,trainingtime,limitedtolocalminimumeasily.TheuseofMATLABtoachieveavarietyofBPneuralnetworktoachievethedesignandtraining,tocomparetheperformanceofBPneuralnetworktoverifytheadvantagesofimprovingtheBPnetwork,howtodrawtheobjectselectedinaccordancewiththeconclusionsofneuralnetworks.Keywords:Artificialneuralnetwork,BPneuralnetworks,errorback-propagationalgorithm,MATLAB,simulation安徽工业大学管理科学与工程学院基于MATLAB神经网络仿真共45页第3页┊┊┊┊┊┊┊┊┊┊┊┊┊装┊┊┊┊┊订┊┊┊┊┊线┊┊┊┊┊┊┊┊┊┊┊┊┊目录1.绪论.................................................................51.1引言.............................................................51.2神经网络概述.....................................................51．2．1神经网络起源...............................................51．2．2神经网络的发展历程.........................................51．2．3神经网络国内发展概况.......................................61．2．4神经网络研究现状...........................................71.3研究目的、方法和问题（BP神经网络）...............................81．3．1研究目的...................................................81．3．2研究方法...................................................81．3．3研究问题...................................................82.BP神经网络.........................................................102.1BP神经网络相关原理..............................................102．1．1神经元非线性模型..........................................102．1．2有教师监督学习............................................102．1．3神经元数学模型............................................112．1．4Delta学习规则.............................................112．1．5神经元激活函数............................................122．1．6BP神经网络收敛准则........................................122.2BP神经网络学习过程描述..........................................132．2．1BP神经网络计算模型建立....................................132．2．2BP神经网络学习过程描述....................................142．2．3BP神经网络方框图..........................................142.3BP神经网络学习方法..............................................142．3．1BP神经网络信号流程........................................142．3．2误差反向传播计算..........................................152．3．3BP神经网络算法描述........................................182.4影响因素分析....................................................192．4．1权值初始值设置影响分析....................................192．4．2权值调整方法影响分析......................................192．4．3激活函数选择影响分析......................................202．4．4学习率η选择影响分析.....................................202．4．5输入输出归一化影响分析....................................212．4．6其他影响因素分析..........................................222.5BP学习算法的改进................................................222．5．1BP学习算法的优缺点........................................222．5．2增加动量项................................................232．5．3弹性BP学习算法...........................................232．5．4自适应学习速率法..........................................24安徽工业大学管理科学与工程学院基于MATLAB神经网络仿真共45页第4页┊┊┊┊┊┊┊┊┊┊┊┊┊装┊┊┊┊┊订┊┊┊┊┊线┊┊┊┊┊┊┊┊┊┊┊┊┊2．5．5共轭梯度法................................................252．5．6Levenberg-Marquardt算法...................................253.BP神经网络仿真.....................................................273.1仿真平台MATLAB..................................................273．1．1MATLAB简介................................................273．1．2仿真平台的构建和策略......................................273.2仿真实验........................................................283．2．1BP神经网络MATLAB设计.....................................283．2．2各种BP学习算法MATLAB仿真................................293．2．3各种算法仿真结果比较与分析................................323．2．4调整初始权值和阈值的仿真..................................333．2．5其他影响因素仿真..........................................354.BP神经网络应用实例.................................................374.1实例概述........................................................374.2网络设计........................................................374.3网络训练........................................................384.4网络测试................................