概率神经网络的matlab源代码

MATLAB程序代码--bp神经网络通用代码matlab通用神经网络代码学习了一段时间的神经网络,总结了一些经验,在这愿意和大家分享一下, 希望对大家有帮助,也希望大家可以把其他神经网络的通用代码在这一起分享感应器神经网络、线性网络、BP神经网络、径向基函数网络%通用感应器神经网络。

P=[-0.5 -0.5 0.3 -0.1 -40;-0.5 0.5 -0.5 1 50];%输入向量T=[1 1 0 0 1];%期望输出plotpv(P,T);%描绘输入点图像net=newp([-40 1;-1 50],1);%生成网络，其中参数分别为输入向量的范围和神经元感应器数量hold onlinehandle=plotpc(net.iw{1},net.b{1});net.adaptparam.passes=3;for a=1:25%训练次数[net,Y,E]=adapt(net,P,T);linehandle=plotpc(net.iw{1},net.b{1},linehandle);drawnow;end%通用newlin程序%通用线性网络进行预测time=0:0.025:5;T=sin(time*4*pi);Q=length(T);P=zeros(5,Q);%P中存储信号T的前5(可变，根据需要而定)次值，作为网络输入。

P(1,2:Q)=T(1,1:(Q-1));P(2,3:Q)=T(1,1:(Q-2));P(3,4:Q)=T(1,1:(Q-3));P(4,5:Q)=T(1,1:(Q-4));P(5,6:Q)=T(1,1:(Q-5));plot(time,T)%绘制信号T曲线xlabel('时间');ylabel('目标信号');title('待预测信号');net=newlind(P,T);%根据输入和期望输出直接生成线性网络a=sim(net,P);%网络测试figure(2)plot(time,a,time,T,'+')xlabel('时间');ylabel('输出-目标+');title('输出信号和目标信号');e=T-a;figure(3)plot(time,e)hold onplot([min(time) max(time)],[0 0],'r:')%可用plot(x,zeros(size(x)),'r:')代替hold offxlabel('时间');ylabel('误差');title('误差信号');%通用BP神经网络P=[-1 -1 2 2;0 5 0 5];t=[-1 -1 1 1];net=newff(minmax(P),[3,1],{'tansig','purelin'},'traingd');%输入参数依次为:'样本P范围',[各层神经元数目],{各层传递函数},'训练函数'%训练函数traingd--梯度下降法，有7个训练参数.%训练函数traingdm--有动量的梯度下降法,附加1个训练参数mc(动量因子，缺省为0.9)%训练函数traingda--有自适应lr的梯度下降法,附加3个训练参数:lr_inc(学习率增长比，缺省为1.05;% lr_dec(学习率下降比，缺省为0.7);max_perf_inc(表现函数增加最大比，缺省为1.04)%训练函数traingdx--有动量的梯度下降法中赋以自适应lr的方法，附加traingdm和traingda的4个附加参数%训练函数trainrp--弹性梯度下降法,可以消除输入数值很大或很小时的误差，附加4个训练参数: % delt_inc(权值变化增加量，缺省为1.2);delt_dec(权值变化减小量，缺省为0.5);% delta0(初始权值变化，缺省为0.07);deltamax(权值变化最大值,缺省为50.0)% 适合大型网络%训练函数traincgf--Fletcher-Reeves共轭梯度法;训练函数traincgp--Polak-Ribiere共轭梯度法;%训练函数traincgb--Powell-Beale共轭梯度法%共轭梯度法占用存储空间小,附加1训练参数searchFcn(一维线性搜索方法,缺省为srchcha);缺少1个训练参数lr%训练函数trainscg--量化共轭梯度法,与其他共轭梯度法相比，节约时间.适合大型网络% 附加2个训练参数:sigma(因为二次求导对权值调整的影响参数，缺省为5.0e-5);% lambda(Hessian阵不确定性调节参数，缺省为5.0e-7)% 缺少1个训练参数:lr%训练函数trainbfg--BFGS拟牛顿回退法,收敛速度快,但需要更多内存,与共轭梯度法训练参数相同,适合小网络%训练函数trainoss--一步正割的BP训练法,解决了BFGS消耗内存的问题,与共轭梯度法训练参数相同%训练函数trainlm--Levenberg-Marquardt训练法,用于内存充足的中小型网络net=init(net);net.trainparam.epochs=300; %最大训练次数(前缺省为10,自trainrp后，缺省为100)net.trainparam.lr=0.05; %学习率(缺省为0.01)net.trainparam.show=50; %限时训练迭代过程(NaN表示不显示，缺省为25)net.trainparam.goal=1e-5; %训练要求精度(缺省为0)%net.trainparam.max_fail 最大失败次数(缺省为5)%net.trainparam.min_grad 最小梯度要求(前缺省为1e-10，自trainrp后，缺省为1e-6) %net.trainparam.time 最大训练时间(缺省为inf)[net,tr]=train(net,P,t); %网络训练a=sim(net,P) %网络仿真%通用径向基函数网络——%其在逼近能力,分类能力,学习速度方面均优于BP神经网络%在径向基网络中,径向基层的散步常数是spread的选取是关键%spread越大,需要的神经元越少,但精度会相应下降,spread的缺省值为1%可以通过net=newrbe(P,T,spread)生成网络,且误差为0%可以通过net=newrb(P,T,goal,spread)生成网络,神经元由1开始增加,直到达到训练精度或神经元数目最多为止%GRNN网络,迅速生成广义回归神经网络(GRNN)P=[4 5 6];T=[1.5 3.6 6.7];net=newgrnn(P,T);%仿真验证p=4.5;v=sim(net,p)%PNN网络,概率神经网络P=[0 0 ;1 1;0 3;1 4;3 1;4 1;4 3]';Tc=[1 1 2 2 3 3 3];%将期望输出通过ind2vec()转换,并设计、验证网络T=ind2vec(Tc);net=newpnn(P,T);Y=sim(net,P);Yc=vec2ind(Y)%尝试用其他的输入向量验证网络P2=[1 4;0 1;5 2]';Y=sim(net,P2);Yc=vec2ind(Y)%应用newrb()函数构建径向基网络,对一系列数据点进行函数逼近P=-1:0.1:1;T=[-0.9602 -0.5770 -0.0729 0.3771 0.6405 0.6600 0.4609...0.1336 -0.2013 -0.4344 -0.500 -0.3930 -0.1647 -0.0988...0.3072 0.3960 0.3449 0.1816 -0.0312 -0.2189 -0.3201];%绘制训练用样本的数据点plot(P,T,'r*');title('训练样本');xlabel('输入向量P');ylabel('目标向量T');%设计一个径向基函数网络，网络有两层,隐层为径向基神经元,输出层为线性神经元%绘制隐层神经元径向基传递函数的曲线p=-3:.1:3;a=radbas(p);plot(p,a)title('径向基传递函数')xlabel('输入向量p')%隐层神经元的权值、阈值与径向基函数的位置和宽度有关,只要隐层神经元数目、权值、阈值正确,可逼近任意函数%例如a2=radbas(p-1.5);a3=radbas(p+2);a4=a+a2*1.5+a3*0.5;plot(p,a,'b',p,a2,'g',p,a3,'r',p,a4,'m--')title('径向基传递函数权值之和')xlabel('输入p');ylabel('输出a');%应用newrb()函数构建径向基网络的时候,可以预先设定均方差精度eg以及散布常数sc eg=0.02;sc=1; %其值的选取与最终网络的效果有很大关系,过小造成过适性,过大造成重叠性net=newrb(P,T,eg,sc);%网络测试plot(P,T,'*')xlabel('输入');X=-1:.01:1;Y=sim(net,X);hold onplot(X,Y);hold offlegend('目标','输出')%应用grnn进行函数逼近P=[1 2 3 4 5 6 7 8];T=[0 1 2 3 2 1 2 1];plot(P,T,'.','markersize',30)axis([0 9 -1 4])title('待逼近函数')xlabel('P')ylabel('T')%网络设计%对于离散数据点,散布常数spread选取比输入向量之间的距离稍小一些spread=0.7;net=newgrnn(P,T,spread);%网络测试A=sim(net,P);hold onoutputline=plot(P,A,'o','markersize',10,'color',[1 0 0]);title('检测网络')xlabel('P')ylabel('T和A')%应用pnn进行变量的分类P=[1 2;2 2;1 1]; %输入向量Tc=[1 2 3]; %P对应的三个期望输出%绘制出输入向量及其相对应的类别plot(P(1,:),P(2,:),'.','markersize',30)for i=1:3text(P(1,i)+0.1,P(2,i),sprintf('class %g',Tc(i)))endaxis([0 3 0 3]);title('三向量及其类别')xlabel('P(1,:)')ylabel('P(2,:)')%网络设计T=ind2vec(Tc);spread=1;net=newgrnn(P,T,speard);%网络测试A=sim(net,P);Ac=vec2ind(A);%绘制输入向量及其相应的网络输出plot(P(1,:),P(2,:),'.','markersize',30)for i=1:3text(P(1,i)+0.1,P(2,i),sprintf('class %g',Ac(i)))endaxis([0 3 0 3]);title('网络测试结果')xlabel('P(1,:)')ylabel('P(2,:)')P=[13, 0, 1.119, 1, 26.3;22, 0, 1.135, 1, 26.3;-15, 0, 0.9017, 1, 20.4;-30, 0, 0.9172, 1, 26.7;24, 0, 1.238,0.9704,28.2;3,24,1.119,1,26.3;0,52,1.089,1,26.3;0,-73,1.0889,1,26.3;1,28, 0.8748,1,26.3;-1,-39,1.1168,1,26.7;-2, 0, 1.495, 1, 26.3;0, -1, 1.438, 1, 26.3;4, 1,0.4964, 0.9021, 26.3;3, -1, 0.5533, 1.2357, 26.7;-5, 0, 1.7368, 1, 26.7;1, 0, 1.1045, 0.0202, 26.3;-2, 0, 1.1168, 1.3764, 26.7;-3, -1, 1.1655, 1.4418,27.5;3, 2, 1.0875, 0.748, 27.5;-3, 0, 1.1068, 2.2092, 26.3;4, 1, 0.9017, 1, 13.7;3, 2, 0.9017, 1, 14.9;-3, 1, 0.9172, 1, 13.7;-2, 0, 1.0198, 1.0809, 16.1;0, 1, 0.9172, 1, 13.7] T=[1, 0, 0, 0, 0 ;1, 0, 0, 0, 0 ;1, 0, 0, 0, 0 ;1, 0, 0, 0, 0 ;1, 0, 0, 0, 0; 0, 1, 0, 0, 0;0, 1, 0, 0, 0;0, 1, 0, 0, 0;0, 1, 0, 0, 0;0, 1, 0, 0, 0;0, 0, 1, 0, 0;0, 0, 1, 0, 0;0, 0, 1, 0, 0;0, 0, 1, 0, 0;0, 0, 1, 0, 0;0, 0, 0, 1, 0 ;0, 0, 0, 1, 0 ;0, 0, 0, 1, 0 ;0, 0, 0, 1, 0 ;0, 0, 0, 1, 0 ; 0, 0, 0, 0, 1;0, 0, 0, 0, 1;0, 0, 0, 0, 1;0, 0, 0, 0, 1;0, 0, 0, 0, 1 ];%期望输出plotpv(P,T);%描绘输入点图像。

Matlab神经网络30个案例第1案例代码%% 清空环境变量clcclear%% 训练数据预测数据提取及归一化%下载四类语音信号load data1 c1load data2 c2load data3 c3load data4 c4%四个特征信号矩阵合成一个矩阵data(1:500,:)=c1(1:500,:);data(501:1000,:)=c2(1:500,:);data(1001:1500,:)=c3(1:500,:);data(1501:2000,:)=c4(1:500,:);%从1到2000间随机排序k=rand(1,2000);[m,n]=sort(k);%输入输出数据input=data(:,2:25);output1 =data(:,1);%把输出从1维变成4维for i=1:2000switch output1(i)case 1output(i,:)=[1 0 0 0];case 2output(i,:)=[0 1 0 0];case 3output(i,:)=[0 0 1 0];case 4output(i,:)=[0 0 0 1];endend%随机提取1500个样本为训练样本，500个样本为预测样本input_train=input(n(1:1500),:)';output_train=output(n(1:1500),:)';input_test=input(n(1501:2000),:)';output_test=output(n(1501:2000),:)';%输入数据归一化[inputn,inputps]=mapminmax(input_train);%% 网络结构初始化innum=24;midnum=25;outnum=4;%权值初始化w1=rands(midnum,innum);b1=rands(midnum,1);w2=rands(midnum,outnum);b2=rands(outnum,1);w2_1=w2;w2_2=w2_1;w1_1=w1;w1_2=w1_1;b1_1=b1;b1_2=b1_1;b2_1=b2;b2_2=b2_1;%学习率xite=0.1alfa=0.01;%% 网络训练for ii=1:10E(ii)=0;for i=1:1:1500%% 网络预测输出x=inputn(:,i);% 隐含层输出for j=1:1:midnumI(j)=inputn(:,i)'*w1(j,:)'+b1(j);Iout(j)=1/(1+exp(-I(j)));end% 输出层输出yn=w2'*Iout'+b2;%% 权值阀值修正%计算误差e=output_train(:,i)-yn;E(ii)=E(ii)+sum(abs(e));%计算权值变化率dw2=e*Iout;db2=e';for j=1:1:midnumS=1/(1+exp(-I(j)));FI(j)=S*(1-S);endfor k=1:1:innumfor j=1:1:midnumdw1(k,j)=FI(j)*x(k)*(e(1)*w2(j,1)+e(2)*w2(j,2)+e(3)*w2(j,3)+ e(4)*w2(j,4));db1(j)=FI(j)*(e(1)*w2(j,1)+e(2)*w2(j,2)+e(3)*w2(j,3)+e(4)*w 2(j,4));endendw1=w1_1+xite*dw1';b1=b1_1+xite*db1';w2=w2_1+xite*dw2';b2=b2_1+xite*db2';w1_2=w1_1;w1_1=w1;w2_2=w2_1;w2_1=w2;b1_2=b1_1;b1_1=b1;b2_2=b2_1;b2_1=b2;endend%% 语音特征信号分类inputn_test=mapminmax('apply',input_test,inputps); for ii=1:1for i=1:500%1500%隐含层输出for j=1:1:midnumI(j)=inputn_test(:,i)'*w1(j,:)'+b1(j);Iout(j)=1/(1+exp(-I(j)));endfore(:,i)=w2'*Iout'+b2;endend%% 结果分析%根据网络输出找出数据属于哪类for i=1:500output_fore(i)=find(fore(:,i)==max(fore(:,i))); end%BP网络预测误差error=output_fore-output1(n(1501:2000))';%画出预测语音种类和实际语音种类的分类图figure(1) plot(output_fore,'r')hold onplot(output1(n(1501:2000))','b') legend('预测语音类别','实际语音类别') %画出误差图figure(2)plot(error)title('BP网络分类误差','fontsize',12) xlabel('语音信号','fontsize',12) ylabel('分类误差','fontsize',12)%print -dtiff -r600 1-4k=zeros(1,4);%找出判断错误的分类属于哪一类for i=1:500if error(i)~=0[b,c]=max(output_test(:,i));switch ccase 1k(1)=k(1)+1;。

30个智能算法matlab代码以下是30个使用MATLAB编写的智能算法的示例代码： 1. 线性回归算法：matlab.x = [1, 2, 3, 4, 5];y = [2, 4, 6, 8, 10];coefficients = polyfit(x, y, 1);predicted_y = polyval(coefficients, x);2. 逻辑回归算法：matlab.x = [1, 2, 3, 4, 5];y = [0, 0, 1, 1, 1];model = fitglm(x, y, 'Distribution', 'binomial'); predicted_y = predict(model, x);3. 支持向量机算法：matlab.x = [1, 2, 3, 4, 5; 1, 2, 2, 3, 3];y = [1, 1, -1, -1, -1];model = fitcsvm(x', y');predicted_y = predict(model, x');4. 决策树算法：matlab.x = [1, 2, 3, 4, 5; 1, 2, 2, 3, 3]; y = [0, 0, 1, 1, 1];model = fitctree(x', y');predicted_y = predict(model, x');5. 随机森林算法：matlab.x = [1, 2, 3, 4, 5; 1, 2, 2, 3, 3]; y = [0, 0, 1, 1, 1];model = TreeBagger(50, x', y');predicted_y = predict(model, x');6. K均值聚类算法：matlab.x = [1, 2, 3, 10, 11, 12]; y = [1, 2, 3, 10, 11, 12]; data = [x', y'];idx = kmeans(data, 2);7. DBSCAN聚类算法：matlab.x = [1, 2, 3, 10, 11, 12]; y = [1, 2, 3, 10, 11, 12]; data = [x', y'];epsilon = 2;minPts = 2;[idx, corePoints] = dbscan(data, epsilon, minPts);8. 神经网络算法：matlab.x = [1, 2, 3, 4, 5];y = [0, 0, 1, 1, 1];net = feedforwardnet(10);net = train(net, x', y');predicted_y = net(x');9. 遗传算法：matlab.fitnessFunction = @(x) x^2 4x + 4;nvars = 1;lb = 0;ub = 5;options = gaoptimset('PlotFcns', @gaplotbestf);[x, fval] = ga(fitnessFunction, nvars, [], [], [], [], lb, ub, [], options);10. 粒子群优化算法：matlab.fitnessFunction = @(x) x^2 4x + 4;nvars = 1;lb = 0;ub = 5;options = optimoptions('particleswarm', 'PlotFcn',@pswplotbestf);[x, fval] = particleswarm(fitnessFunction, nvars, lb, ub, options);11. 蚁群算法：matlab.distanceMatrix = [0, 2, 3; 2, 0, 4; 3, 4, 0];pheromoneMatrix = ones(3, 3);alpha = 1;beta = 1;iterations = 10;bestPath = antColonyOptimization(distanceMatrix, pheromoneMatrix, alpha, beta, iterations);12. 粒子群-蚁群混合算法：matlab.distanceMatrix = [0, 2, 3; 2, 0, 4; 3, 4, 0];pheromoneMatrix = ones(3, 3);alpha = 1;beta = 1;iterations = 10;bestPath = particleAntHybrid(distanceMatrix, pheromoneMatrix, alpha, beta, iterations);13. 遗传算法-粒子群混合算法：matlab.fitnessFunction = @(x) x^2 4x + 4;nvars = 1;lb = 0;ub = 5;gaOptions = gaoptimset('PlotFcns', @gaplotbestf);psOptions = optimoptions('particleswarm', 'PlotFcn',@pswplotbestf);[x, fval] = gaParticleHybrid(fitnessFunction, nvars, lb, ub, gaOptions, psOptions);14. K近邻算法：matlab.x = [1, 2, 3, 4, 5; 1, 2, 2, 3, 3]; y = [0, 0, 1, 1, 1];model = fitcknn(x', y');predicted_y = predict(model, x');15. 朴素贝叶斯算法：matlab.x = [1, 2, 3, 4, 5; 1, 2, 2, 3, 3]; y = [0, 0, 1, 1, 1];model = fitcnb(x', y');predicted_y = predict(model, x');16. AdaBoost算法：matlab.x = [1, 2, 3, 4, 5; 1, 2, 2, 3, 3];y = [0, 0, 1, 1, 1];model = fitensemble(x', y', 'AdaBoostM1', 100, 'Tree'); predicted_y = predict(model, x');17. 高斯混合模型算法：matlab.x = [1, 2, 3, 4, 5]';y = [0, 0, 1, 1, 1]';data = [x, y];model = fitgmdist(data, 2);idx = cluster(model, data);18. 主成分分析算法：matlab.x = [1, 2, 3, 4, 5; 1, 2, 2, 3, 3]; coefficients = pca(x');transformed_x = x' coefficients;19. 独立成分分析算法：matlab.x = [1, 2, 3, 4, 5; 1, 2, 2, 3, 3]; coefficients = fastica(x');transformed_x = x' coefficients;20. 模糊C均值聚类算法：matlab.x = [1, 2, 3, 4, 5; 1, 2, 2, 3, 3]; options = [2, 100, 1e-5, 0];[centers, U] = fcm(x', 2, options);21. 遗传规划算法：matlab.fitnessFunction = @(x) x^2 4x + 4; nvars = 1;lb = 0;ub = 5;options = optimoptions('ga', 'PlotFcn', @gaplotbestf);[x, fval] = ga(fitnessFunction, nvars, [], [], [], [], lb, ub, [], options);22. 线性规划算法：matlab.f = [-5; -4];A = [1, 2; 3, 1];b = [8; 6];lb = [0; 0];ub = [];[x, fval] = linprog(f, A, b, [], [], lb, ub);23. 整数规划算法：matlab.f = [-5; -4];A = [1, 2; 3, 1];b = [8; 6];intcon = [1, 2];[x, fval] = intlinprog(f, intcon, A, b);24. 图像分割算法：matlab.image = imread('image.jpg');grayImage = rgb2gray(image);binaryImage = imbinarize(grayImage);segmented = medfilt2(binaryImage);25. 文本分类算法：matlab.documents = ["This is a document.", "Another document.", "Yet another document."];labels = categorical(["Class 1", "Class 2", "Class 1"]);model = trainTextClassifier(documents, labels);newDocuments = ["A new document.", "Another new document."];predictedLabels = classifyText(model, newDocuments);26. 图像识别算法：matlab.image = imread('image.jpg');features = extractFeatures(image);model = trainImageClassifier(features, labels);newImage = imread('new_image.jpg');newFeatures = extractFeatures(newImage);predictedLabel = classifyImage(model, newFeatures);27. 时间序列预测算法：matlab.data = [1, 2, 3, 4, 5];model = arima(2, 1, 1);model = estimate(model, data);forecastedData = forecast(model, 5);28. 关联规则挖掘算法：matlab.data = readtable('data.csv');rules = associationRules(data, 'Support', 0.1);29. 增强学习算法：matlab.environment = rlPredefinedEnv('Pendulum');agent = rlDDPGAgent(environment);train(agent);30. 马尔可夫决策过程算法：matlab.states = [1, 2, 3];actions = [1, 2];transitionMatrix = [0.8, 0.1, 0.1; 0.2, 0.6, 0.2; 0.3, 0.3, 0.4];rewardMatrix = [1, 0, -1; -1, 1, 0; 0, -1, 1];policy = mdpPolicyIteration(transitionMatrix, rewardMatrix);以上是30个使用MATLAB编写的智能算法的示例代码，每个算法都可以根据具体的问题和数据进行相应的调整和优化。

神经网络及深度学习(包含matlab代码).pdf
神经网络可以使用中间层构建出多层抽象，正如在布尔电路中所做的那样。

如果进行视觉模式识别，那么第1 层的神经元可能学会识别边；第2 层的神经元可以在此基础上学会识别更加复杂的形状，例如三角形或矩形；第3 层将能够识别更加复杂的形状，以此类推。

有了这些多层抽象，深度神经网络似乎可以学习解决复杂的模式识别问题。

正如电路示例所体现的那样，理论研究表明深度神经网络本质上比浅层神经网络更强大。

《深入浅出神经网络与深度学习》PDF+代码分析
《深入浅出神经网络与深度学习》PDF中文，249页；PDF英文，292页；配套代码。

提取码: 6sgh
以技术原理为导向，辅以MNIST 手写数字识别项目示例，介绍神经网络架构、反向传播算法、过拟合解决方案、卷积神经网络等内容，以及如何利用这些知识改进深度学习项目。

学完后，将能够通过编写Python 代码来解决复杂的模式识别问题。

BP神经网络的设计MATLAB编程例1 采用动量梯度下降算法训练 BP 网络。

训练样本定义如下：输入矢量为p =[-1 -2 3 1-1 1 5 -3]目标矢量为 t = [-1 -1 1 1]解：本例的 MATLAB 程序如下：close allclearecho onclc% NEWFF——生成一个新的前向神经网络% TRAIN——对 BP 神经网络进行训练% SIM——对 BP 神经网络进行仿真pause% 敲任意键开始clc% 定义训练样本% P 为输入矢量P=[-1, -2, 3, 1; -1, 1, 5, -3];% T 为目标矢量T=[-1, -1, 1, 1];pause;clc% 创建一个新的前向神经网络net=newff(minmax(P),[3,1],{'tansig','purelin'},'traingdm')% 当前输入层权值和阈值inputWeights=net.IW{1,1}inputbias=net.b{1}% 当前网络层权值和阈值layerWeights=net.LW{2,1}layerbias=net.b{2}pauseclc% 设置训练参数net.trainParam.show = 50;net.trainParam.lr = 0.05;net.trainParam.mc = 0.9;net.trainParam.epochs = 1000;net.trainParam.goal = 1e-3;pauseclc% 调用 TRAINGDM 算法训练 BP 网络[net,tr]=train(net,P,T);pauseclc% 对 BP 网络进行仿真A = sim(net,P)% 计算仿真误差E = T - AMSE=mse(E)pauseclcecho off例2 采用贝叶斯正则化算法提高 BP 网络的推广能力。

用matlab编的神经网络程序一、%bp.m-Implementation of backpropagation algorithm%(C)copyright2001by Yu Hen Hu%created:3/17/2001%call bpconfig.m,cvgtest.m,bpdisplay.m,bptest.m%rsample.m,randomize.m,actfun.m,actfunp.m%partunef.m%modified:12/10/2003handle case when testing file has no labelclear all,bpconfig;%configurate the MLP network and learning parameters.%BP iterations beginswhile not_converged==1,%start a new epoch%Randomly select K training samples from the training set.[train,ptr,train0]=rsample(train0,K,Kr,ptr);%train is K by M+Nz{1}=(train(:,1:M))';%input sample matrix M by Kd=train(:,M+1:MN)';%corresponding target value N by K%Feed-forward phase,compute sum of square errorsfor l=2:L,%the l-th layeru{l}=w{l}*[ones(1,K);z{l-1}];%u{l}is n(l)by Kz{l}=actfun(u{l},atype(l));enderror=d-z{L};%error is N by KE(t)=sum(sum(error.*error));%Error back-propagation phase,compute delta errordelta{L}=actfunp(u{L},atype(L)).*error;%N(=n(L))by Kif L>2,for l=L-1:-1:2,delta{l}=(w{l+1}(:,2:n(l)+1))'*delta{l+1}.*actfunp(u{l},atype(l));endend%update the weight matrix using gradient,momentum and random perturbation for l=2:L,dw{l}=alpha*delta{l}*[ones(1,K);z{l-1}]'+...mom*dw{l}+randn(size(w{l}))*0.005;w{l}=w{l}+dw{l};end%display the training errorbpdisplay;%Test convergence to see if the convergence condition is satisfied,cvgtest;t=t+1;%increment epoch countend%while loopdisp('Final training results:')if classreg==0,[Cmat,crate]=bptest(wbest,tune,atype),elseif classreg==1,SS=bptestap(wbest,tune,atype),endif testys==1,disp('Apply trained MLP network to the testing data.The results are:');if classreg==0,[Cmat,crate]=bptest(wbest,test0,atype),elseif classreg==1,[SS,out]=bptestap(wbest,test0,atype);figure(2),clf,plot(test0,out),title('output of testing results')endend二、echo on%LearnER.m-Example of Error Correcting Learning%copyright(c)1996-2000by Yu Hen Hu%Created:9/2/96%Modified:1/28/2000%Modified:9/3/2001add additional runs of LMS and display weight converge curve %clear allx=[11110.5-0.41.10.70.80.4-0.31.2];d=[1001];w=zeros(3,1);weight=[];eta=.01;echo off;pausefor n=1:4,y(n)=w'*x(:,n);e(n)=d(n)-y(n);w=w+eta*e(n)*x(:,n);weight=[weight w];['iteration#'int2str(n)':']weightpauseendfor m=1:499,x0=randomize(x')';%change the order of presentation of x for n=1:4,y(n)=w'*x0(:,n);e(n)=d(n)-y(n);w=w+eta*e(n)*x0(:,n);weight=[weight w];endendfigure(1),subplot(311),plot([1:size(weight,2)],weight(1,),ylabel('w0') title('convergence curve of the weights')subplot(312),plot([1:size(weight,2)],weight(2,),ylabel('w1') subplot(313),plot([1:size(weight,2)],weight(3,),ylabel('w2')echo on%Batched mode LS solutionR=x*x'rho=sum((x*diag(d))')'w_ls=inv(R)*rhoerror=d-w_ls'*x;ernorm=error*error'echo off三、%perceptron.m-perceptron learning algorithm%INput:train(:,1:M)-pattern train(:,M+1)-target%Output:weight vector w=[w0w1...wM],w0:bias%actual output vector y%Need to call m-file routine:datasepf.m,sline.m%copyright(C)1996-2001by Yu Hen Hu%Modified:2/9/2000,2/3/2001%K2:#of training samples%M:feature space dimensionclear all,clfgdata=input('Enter0to load a data file,Return to generate separable data:');if isempty(gdata)|gdata~=0,%generate random training dataK2=input('number of training samples=');[orig,slope]=datasepf(K2);%slope is the slope of separating plane%that has the formula:y=slope*x+0.5*(1-slope)elsedisp('enter the data matrix,row by row,[x1..xN t]');orig=input('start from class1,followed by class0:');end[Km,Kn]=size(orig);M=Kn-1;%number of inputsK0=sum([orig(:,Kn)==0]);K1=Km-K0;K2=K0+K1;%#of targets=0and1 mdisplay=10;%#of displaying hyperplane before checking for stopping%Initial hyperplane%w=[rand(1,M)0];%initial random weights%The initial hyperplane can be estimated as a hyperplane separating%a pair of data sample with different labels%in orig,this is the first and the last data sample since there are%only two classes and sorted according to class labels%the separating hyperplane of two points a and b%has the normal vector[-0.5(|b|^2-|a|^2)b-a]wa=orig(1,1:M);wb=orig(K2,1:M);wmag=0.5*(wb*wb'-wa*wa');wunit=wb-wa;w=[-wmag wunit(1)wunit(2)];figure(1)subplot(1,2,2),plot(orig(1:K1,1),orig(1:K1,2),'*',orig(K1+1:K2,1),orig(K1+1:K2,2),'o') axis('square');v=axis;[lx,ly]=sline(w,v);subplot(1,2,1),plot(orig(1:K1,1),orig(1:K1,2),'*',...orig(K1+1:K2,1),orig(K1+1:K2,2),'o',lx,ly)axis('square');title('Initial hyperplane')converged=0;%0<eta<1/x(k)_max.etamax=sqrt(max(orig(:,1).*orig(:,1)+orig(:,2).*orig(:,2))); eta=input(['0<eta<'num2str(etamax)',Enter eta=']) epoch=0;while converged==0,%not converged yettrain=randomize(orig);for i=1:K2,y(i)=0.5*(1+sign(w*[1;train(i,1:M)']));w=w+eta*(train(i,M+1)-y(i))*[1train(i,1:M)];[lx,ly]=sline(w,v);subplot(1,2,2),plot(orig(1:K1,1),orig(1:K1,2),'*g',...orig(K1+1:K2,1),orig(K1+1:K2,2),'og',lx,ly,'-',...train(i,1),train(i,2),'sr');axis('square');pause(0.1)drawnowend%for loopepoch=epoch+1;if sum(abs(train(:,M+1)-y'))==0,%check if converged converged=1;endif rem(epoch,mdisplay)==0,converged=input('type1to terminate,Return to continue:') if isempty(converged),converged=0;endendif converged==1,[lx,ly]=sline(w,v);subplot(1,2,2),plot(orig(1:K1,1),orig(1:K1,2),'*',...orig(K1+1:K2,1),orig(K1+1:K2,2),'o',lx,ly)axis('square');title('final hyperplane location')endend%while loop。

本人编辑了一个预测模型的程序，但是matlab老是提示出错，请各位大虾指教：非常感谢！！！>> clear all;%define the input and outputp= [974 874 527;388 466 1764;1316 2439 2251;1836 2410 1860;1557 2301 1578;1490 1877 2749;1513 1278 2026;1070 1561 2794;1347 2415 3306;1324 2746 1233;1383 1463 1847;1282 0 2347];t=[19797;24282;34548];% 创建bp网络和定义训练函数net=newff([388 3306],[15 1],{'tansig' 'purelin'});net.trainparam.goal=50;net.trainparam.epochs=5000;%训练神经网络[net,tr]=train(net,p,t);%输出训练后的权值和阈值iw1=net.IW{1};b1=net.b{1};lw2=net.LW{2};b2=net.b{2};%存储训练好的神经网络save netkohler net??? Error using ==> network.trainInputs are incorrectly sized for network.Matrix must have 1 rows.修改后：>> clear all;%define the input and outputp= [974 874 527;388 466 1764;1316 2439 2251;1836 2410 1860;1557 2301 1578;1490 1877 2749;1513 1278 2026;1070 1561 2794;1347 2415 3306;1324 2746 1233;1383 1463 1847;1282 0 2347];t=[19797;24282;34548];% 创建bp网络和定义训练函数net=newff([388 1836;466 2746;527 3306],[15 1],{'tansig' 'purelin'});net.trainparam.goal=50;net.trainparam.epochs=5000;%训练神经网络[net,tr]=train(net,p,t);%输出训练后的权值和阈值iw1=net.IW{1};b1=net.b{1};lw2=net.LW{2};b2=net.b{2};%存储训练好的神经网络save netkohler net??? Error using ==> network.trainInputs are incorrectly sized for network.Matrix must have 3 rows.clear all;%define the input and outputp= [974 874 527;388 466 1764;1316 2439 2251;1836 2410 1860;1557 2301 1578;1490 1877 2749;1513 1278 2026;1070 1561 2794;1347 2415 3306;1324 2746 1233;1383 1463 1847;1282 0 2347];t=[19797 24282 34548];% 创建bp网络和定义训练函数pr=[527 974;388 1764;1316 2439;1836 2410;1557 2301;1490 2749;％这里是为了方便而建立一个矩阵，注意是12x2，不是3x21278 2026;1070 2794;1347 3306;1233 2746;1383 1847;0 2347]net=newff(pr,[15,1],{'tansig' 'purelin'},'trainlm');％这里要加入输出层的转移函数，一般是trainlm net.trainparam.goal=50;net.trainparam.epochs=5000;%训练神经网络[net,tr]=train(net,p,t);%输出训练后的权值和阈值iw1=net.IW{1};b1=net.b{1};lw2=net.LW{2};b2=net.b{2};%存储训练好的神经网络save netkohler net。

MATLAB程序代码--bp神经网络应用举例1BP神经网络的设计实例例1采用动量梯度下降算法训练BP网络。

训练样本定义如下：输入矢量为p=[-1-231-115-3]目标矢量为t=[-1-111]解：本例的MATLAB程序如下：close allclearecho onclc%NEWFF——生成一个新的前向神经网络%TRAIN——对BP神经网络进行训练%SIM——对BP神经网络进行仿真pause%敲任意键开始clc%定义训练样本%P为输入矢量P=[-1,-2,3,1;-1,1,5,-3];%T为目标矢量T=[-1,-1,1,1];pause;clc%创建一个新的前向神经网络net=newff(minmax(P),[3,1],{'tansig','purelin'},'traingdm')%当前输入层权值和阈值inputWeights=net.IW{1,1}inputbias=net.b{1}%当前网络层权值和阈值layerWeights=net.LW{2,1}layerbias=net.b{2}pauseclc%设置训练参数net.trainParam.show=50;%两次显示之间的训练次数,缺省值为25 net.trainParam.lr=0.05;%学习速率net.trainParam.mc=0.9;%动量常数设置,缺省就是0.9net.trainParam.epochs=1000;%训练次数,缺省值为100net.trainParam.goal=1e-3;%网络性能目标,缺省值为0clc%调用TRAINGDM算法训练BP网络[net,tr]=train(net,P,T);pauseclc%对BP网络进行仿真A=sim(net,P)%计算仿真误差E=T-AMSE=mse(E)pauseclcecho off例2采用贝叶斯正则化算法提高BP网络的推广能力。