模式识别_神经网络识别MATLAB实现.doc

《模式识别_神经网络识别MATLAB实现.doc》由会员上传分享，免费在线阅读，更多相关内容在应用文档-天天文库。

1、模糊神经网络模式识别functionretstr=FnnTrain(dt,ld,tt,sp)retstr=-1;%%%%输入参数赋值开始%%%%%%%%%%%%%%%%%%%%%%%%方便调试程序用，程序调试时去掉这部分的注释%dt=4;     %学习阈值%ld=0.05;   %学习进度%tt=10;     %训练次数%sp='datasample.txt';   %一个点的监测数据%%%%输入参数赋值结束%%%%%%%%%%%%%%%%%%%%%%%%globalrecordDimention;globa

2、lsampleNumber;globalweightNumber;globaldistanceThread;globalWW;globallearningDegree;globalepochsNumber;distanceThread=dt;learningDegree=ld;traintimes=tt;A=load(sp);  [ArowAcol]=size(A);%样本个数sampleNumber=Arow;recordDimention=Acol;disp(sampleNumber);WW=A(1,:);WW

3、=[WW[1]];weightNumber=1;epochsNumber=1;  forjj=2:1:sampleNumber     TrainNN2(A(jj,:));endforjt=1:traintimes-1     forjt2=1:sampleNumber         TrainNN2(A(jj,:));     endend%将训练结果写入权值文件dlmwrite('dataw.dat',WW,'t');%%训练子函数%functionTrainNN2(a)globalrecordDimen

4、tion;globalsampleNumber;globalweightNumber;globaldistanceThread;globalWW;globallearningDegree;globalepochsNumber;Ldistance=zeros(2,weightNumber);forj1=1:weightNumber     Ldistance(2,j1)=j1;end%%%%%%输入输出空间的模糊分割forj2=1:weightNumber     Lx=0;     forj3=1:recordDi

5、mention         Lx=Lx+(a(j3)-WW(j2,j3)).*(a(j3)-WW(j2,j3));     end     Ldistance(1,j2)=sqrt(Lx);end%%%%%%计算模糊空间的距离  Lx1=Ldistance(1,1);Lx2=Ldistance(2,1);forj4=2:weightNumber     if(Ldistance(1,j4)

6、;     endend%%%%%%修正模糊规则updateW=0;nowWeight=Lx2;if(Lx1<=distanceThread)     forj6=1:recordDimention         WW(nowWeight,j6)=WW(nowWeight,j6)+learningDegree.*(a(j6)-WW(nowWeight,j6));     end     WW(nowWeight,recordDimention+1)=WW(nowWeight,recordDimention+1)+

7、1;     updateW=1;end    if(updateW==0)     weightNumber=weightNumber+1;     a=[a[1]];     WW=[WW;a];endL-M神经网络算法之MATLAB实现1%此为BP网络训练程序~k>

8、H]x functionretstr=LmTrain(ModelNo,NetPara,TrainPara,InputFun,OutputFun,DataDir)$NiFfBf NNTWARNOFF~TVc'o> retstr=-1

9、;(`l/A+D] %%%%输入参数赋值开始%%%%%%%%%%%%%%%%%%%%%%%DH#D,zNI0 %方便调试程序用，程序调试时去掉这部分的注释T?u:'7n~W# ModelNo='1';;w.9{* NetPara(1)=7;eF*g}`Yib NetPara(2)=1;RVgB.dFf NetPara(3)=6;*dl

10、ep