myrbfnet.m

matlab写的神经网络的几个演示程序

%RBF神经网络实现（聚类）
clc
close all
clear all

InDim=2;%样本输入维数
OutDim=3;% 样本输出维数
figure
title('训练样本');echo off
axis([-2,2,-2,2]);axis on;grid
xlabel('SamIn x');
ylabel('SamIn y');
line([-1 1],[1 1])
line([1 -1],[1 0])
line([-1 -1],[0 1])
line([-1 1],[-0.5 -0.5])
line([-1 1],[-1.5 -1.5])
line([1 1],[-0.5 -1.5])
line([-1 -1],[-0.5 -1.5])
hold on
SamNum=200;%训练样本数
%rand('state',sum(100*clock))
SamIn=(rand(2,SamNum)-0.5)*4;% 随机产生200个[-2,2]区间样本输入
SamOut=[];
for i=1:SamNum
    Sam=SamIn(:,i);
    x=Sam(1,1);
    y=Sam(2,1);
    if((x>-1)&(x<1))==1
        if ((y>x/2+1/2)&(y<1))==1
            plot(x,y,'r+')
            class=[0 1 0]';
        elseif((y<-0.5)&(y>-1.5))==1
            plot(x,y,'rs')
            class=[0 0 1]';
        else
            plot(x,y,'ro')
            class=[1 0 0]';
        end
    else
        plot(x,y,'ro')
        class=[1 0 0]';
    end
    SamOut=[SamOut class];                  %得到样本对应的类别属性
end
MaxEpochs=1000;%最大训练次数
HiddenUnitNum = 30; % 隐节点数，即聚类样本数
Overlap = 1.0; % 隐节点重叠系数

Centers = SamIn(:,1:HiddenUnitNum);
% NumberInClusters = zeros(HiddenUnitNum,1); % 各类中的样本数，初始化为零
% IndexInClusters = zeros(HiddenUnitNum,SamNum); % 各类所含样本的索引号

% while 1
for EndEpochs=1:MaxEpochs
    NumberInClusters = zeros(HiddenUnitNum,1); % 各类中的样本数，初始化为零
    IndexInClusters = zeros(HiddenUnitNum,SamNum); % 各类所含样本的索引号

    % 按最小距离原则对所有样本进行分类
    for i = 1:SamNum
        AllDistance = dist(Centers',SamIn(:,i));
        [MinDist,Pos] = min(AllDistance);
        NumberInClusters(Pos) = NumberInClusters(Pos) + 1;
        IndexInClusters(Pos,NumberInClusters(Pos)) = i;
    end
    % 保存旧的聚类中心
    OldCenters = Centers;
    for i = 1:HiddenUnitNum
        Index = IndexInClusters(i,1:NumberInClusters(i));
        Centers(:,i) = mean(SamIn(:,Index)')';
    end
    % 判断新旧聚类中心是否一致，是则结束聚类
    EqualNum = sum(sum(Centers==OldCenters));
    if EqualNum == InDim*HiddenUnitNum
        break;
    end
end

% 计算各隐节点的扩展常数（宽度）
AllDistances = dist(Centers',Centers); % 计算隐节点数据中心间的距离（矩阵）
Maximum = max(max(AllDistances)); % 找出其中最大的一个距离
for i = 1:HiddenUnitNum % 将对角线上的0 替换为较大的值
    AllDistances(i,i) = Maximum+1;
end
Spreads = Overlap*min(AllDistances)'; % 以隐节点间的最小距离作为扩展常数

% 计算各隐节点的输出权值
Distance = dist(Centers',SamIn); % 计算各样本输入离各数据中心的距离
SpreadsMat = repmat(Spreads,1,SamNum);
HiddenUnitOut = radbas(Distance./SpreadsMat); % 计算隐节点输出阵
HiddenUnitOutEx = [HiddenUnitOut' ones(SamNum,1)]'; % 考虑偏移
W2Ex = SamOut*pinv(HiddenUnitOutEx); % 求广义输出权值
W2 = W2Ex(:,1:HiddenUnitNum); % 输出权值
B2 = W2Ex(:,HiddenUnitNum+1); % 偏移

% TestSamNum=SamNum;
% TestSamIn=SamIn;
TestSamNum=500;% 测试样本数
TestSamIn=(rand(2,TestSamNum)-0.5)*4;

TestDistance = dist(Centers',TestSamIn);
TestSpreadsMat = repmat(Spreads,1,TestSamNum);
TestHiddenUnitOut = radbas(TestDistance./TestSpreadsMat);
B=repmat(B2,1,TestSamNum);
TestNNOut = W2*TestHiddenUnitOut+B;
[val nnclass]=max(TestNNOut);

figure
title('测试结果');echo off
axis([-2,2,-2,2]);axis on;grid
xlabel('Input x');
ylabel('Output y');
line([-1 1],[1 1]);
line([1 -1],[1 0]);
line([-1 -1],[0 1]);
line([-1 1],[-0.5 -0.5]);
line([-1 1],[-1.5 -1.5]);
line([1 1],[-0.5 -1.5]);
line([-1 -1],[-0.5 -1.5]);
hold on

TestSamOut = [];
for i = 1:TestSamNum
    x = TestSamIn(1,i);
    y = TestSamIn(2,i);
    if nnclass(i)==1
        plot(x,y,'ro');
    elseif nnclass(i)==2
        plot(x,y,'r+');
    else
        plot(x,y,'rs');
    end
    if((x>-1)&(x<1))==1
        if ((y>x/2+1/2)&(y<1))==1
            class = 2;
        elseif((y<-0.5)&(y>-1.5))==1
            class = 3;
        else
            class = 1;
        end
    else
        class = 1;
    end
    TestSamOut = [TestSamOut class];
end
Result = ~abs(nnclass-TestSamOut);       % 正确分类显示为1
Percent = sum(Result)/length(Result)     % 正确分类率