用matlab神经网络实现非线性识别

%---------------------------------------------% % % % 工作室提供代做matlab仿真 % % % % 详情请访问：http://cn.mikecrm.com/5k6v1DP % % % %---------------------------------------------% clear all close all clc rand('seed',2); nSampDim=2; %样本是二维的 t1=rand(1000,1).*pi; t2=rand(1000,1).*pi+pi; r=3*rand(1000,1)+5; x1=r.*cos(t1); y1=r.*sin(t1)-1; x2=r.*cos(t2)+5; y2=r.*sin(t2)+4; %p1=[x1,y1]*[cos(pi/50),sin(pi/50);-sin(pi/50),cos(pi/50)]; %p2=[x2,y2]*[cos(pi/50),sin(pi/50);-sin(pi/50),cos(pi/50)]; %A1=p1(:,1); %旋转后x轴的坐标 %A2=p1(:,2); %旋转后y轴的坐标 %C1=p2(:,1); %C2=p2(:,2); A=[x1,y1]; C=[x2,y2]; %AB各取150个进行训练 train_num_A=750; train_num_C=750; nTrainNum=train_num_A+train_num_C; NUM_A=1000; NUM_C=1000; % A随机选取训练与测试的数据 r=randperm(NUM_A); %随机打乱A序列 traind(1:train_num_A,:)=A(r(1:train_num_A),:); %选A中的150个进行训练 testd(1:NUM_A-train_num_A,:)=A(r(train_num_A+1:NUM_A),:); %选A中剩下的150个进行测试 %C随机选取训练与测试的数据 r=randperm(NUM_C); %随机打乱C的序列 traind(train_num_A+1:train_num_A+train_num_C,:)=C(r(1:train_num_C),:); testd(NUM_A-train_num_A+1:NUM_A+NUM_C-train_num_A-train_num_C,:)=C(r(train_num_C+1:NUM_C),:); %赋值 train1=zeros(1,train_num_A+train_num_C); train1(1:train_num_A)=1; test1=zeros(1,NUM_A+NUM_C-train_num_A-train_num_C); test1(1:NUM_A-train_num_A)=1; %%构造网络 net.nIn=2; %两个输入层节点 net.nHidden=10; %3个隐含层节点 net.nOut=1; %一个输出层节点 w=2*(rand(net.nHidden,net.nIn)-1/2); b=2*(rand(net.nHidden,1)-1/2); net.w1=[w,b]; W=2*(rand(net.nOut,net.nHidden)-1/2); B=2*(rand(net.nOut,1)-1/2); net.w2=[W,B]; %%训练数据归一化 %mm=mean(traind); %均值平移 %traind_s=zeros(nTrainNum,2); %for i=1:2 %traind_s(:,i)=traind(:,i)-mm(i); %end %方差标准化 %m1(1)=std(traind_s(:,1)); %m1(2)=std(traind_s(:,2)); %for i=1:2 %traind_s(:,i)=traind_s(:,i)/m1(i); %end %%训练 SampleInEx=[traind';ones(1,nTrainNum)]; expectedOut=train1; eb=0.01; %误差容限 eta=0.1; %学习率 mc=0.1; %动量因子 maxiter=5000; %最大迭代次数 iteration=0; %第一代 errRec=zeros(1,maxiter); outRec=zeros(nTrainNum,maxiter); NET=[];%记录 %开始迭代 for i=1:maxiter hid_input=net.w1*SampleInEx; %隐含层的输入 hid_out=sigmoid(hid_input); %隐含层的输出 ou_input1=[hid_out;ones(1,nTrainNum)]; %输出层的输入 ou_input2=net.w2*ou_input1; out_out=sigmoid(ou_input2); outRec(:,i)=out_out'; err=expectedOut-out_out; %误差 sse=sumsqr(err); errRec(i)=sse; %保存误差值 fprintf('第 %d 次迭代 误差： %f\n', i, sse); iteration=iteration + 1; %判断是否收敛 if sse<=eb break; end %误差反向传播 %隐含层与输出层之间的局部梯度 DELTA=err.*sigmoid(ou_input2).*(1-sigmoid(ou_input2)); %输入层与隐层之间的局部梯度 delta=net.w2(:,1:end-1)'*DELTA.*sigmoid(hid_input).*(1-sigmoid(hid_input)); %权值修改量 dWEX=DELTA*ou_input1'; dwex=delta*SampleInEx'; %修改权值，如果不是第一次修改，则使用动量因子 if i==1 net.w2=net.w2+eta*dWEX; net.w1=net.w1+eta*dwex; else net.w2=net.w2+(1-mc)*eta*dWEX+mc*dWEXOld; net.w1=net.w1+(1-mc)*eta*dwex+mc*dwexOld; end %记录上一次的权值修改量 dWEXOld=dWEX; dwexOld=dwex; end %%测试 %测试数据归一化 %for i=1:2 %testd_s(:,i)=testd(:,i)-mm(i); %end %for i=1:2 %testd_s(:,i)=testd_s(:,1)/m1(i); %end %计算测试输出 InEx=[testd';ones(1,2000-nTrainNum)]; hid_input=net.w1*InEx; hid_out=sigmoid(hid_input); ou_input1=[hid_out;ones(1,2000-nTrainNum)]; ou_input2=net.w2*ou_input1; out_out=sigmoid(ou_input2); outout1=out_out; %取整 out_out(out_out<0.5)=0; out_out(out_out>=0.5)=1; %正确率 rate=sum(out_out==test1)/length(out_out); %%显示 %显示训练样本 train_A=traind(train1==1,:); train_A=train_A'; train_C=traind(train1==0,:); train_C=train_C'; x=-15:0.1:25; [xx,yy]=meshgrid(x); for i=1: length(xx) for j=1:length(yy) xi=[xx(i,j),yy(i,j),1]; hp=net.w1*xi'; tau=sigmoid(hp); tauex=[tau',1]'; HM=net.w2*tauex; out=sigmoid(HM); z(i,j)=out; end end figure(3) plot(train_A(1,:),train_A(2,:),'bo'); hold on; plot(train_C(1,:),train_C(2,:),'r*'); [c,h]=contour(x,x,z,0.5,'b'); clabel(c,h); xlabel('x') ylabel('y') axis([-15 25 -15 25]) title('训练样本分布') legend('A','C') M(i)=getframe; figure(2) mesh(x,x,z); figure(5) plot(x1,y1,'k*') hold on; plot(x2,y2,'ro') figure(4) axis on hold on grid [nRow,nCol]=size(errRec); plot(1:nCol,errRec,'b-','LineWidth',1.5); legend('误差平方和') xlabel('迭代次数','FontName','Times','FontSize',10); ylabel('误差') fprintf('-------错误分类表----------\n') fprintf(' 编号 标签 x y\n') ind= find(out_out ~=test1); for i=1:length(ind) fprintf(' %d %d %f %f \n', ind(i),test1(ind(i)), testd(ind(i),1), testd(ind(i),2)); end fprintf('最终迭代次数\n %d\n', iteration); fprintf('正确率：\n %f%%\n',rate*100);