共有七個完整算法 % 1.基于聚類的RBF 網(wǎng)設(shè)計算法 % 一維輸入，一維輸出，逼近效果很好！ SamNum = 100; % 總樣本數(shù) TestSamNum = 101; % 測試樣本數(shù) InDim = 1; % 樣本輸入維數(shù) ClusterNum = 10; % 隱節(jié)點數(shù)，即聚類樣本數(shù) Overlap = 1.0; % 隱節(jié)點重疊系數(shù) % 根據(jù)目標函數(shù)獲得樣本輸入輸出 rand('state',sum(100*clock)) NoiseVar = 0.1; Noise = NoiseVar*randn(1,SamNum); SamIn = 8*rand(1,SamNum)-4; SamOutNoNoise = 1.1*(1-SamIn+2*SamIn.^2).*exp(-SamIn.^2/2); SamOut = SamOutNoNoise + Noise; TestSamIn = -4:0.08:4; TestSamOut = 1.1*(1-TestSamIn+2*TestSamIn.^2).*exp(-TestSamIn.^2/2); figure hold on grid plot(SamIn,SamOut,'k+') plot(TestSamIn,TestSamOut,'r--') xlabel('Input x'); ylabel('Output y'); Centers = SamIn(:,1:ClusterNum); NumberInClusters = zeros(ClusterNum,1); % 各類中的樣本數(shù)，初始化為零 IndexInClusters = zeros(ClusterNum,SamNum); % 各類所含樣本的索引號 while 1, NumberInClusters = zeros(ClusterNum,1); % 各類中的樣本數(shù)，初始化為零 IndexInClusters = zeros(ClusterNum,SamNum); % 各類所含樣本的索引號 % 按最小距離原則對所有樣本進行分類 for i = 1:SamNum AllDistance = dist(Centers',SamIn(:,i)); [MinDist,Pos] = min(AllDistance); NumberInClusters(Pos) = NumberInClusters(Pos) + 1; IndexInClusters(Pos,NumberInClusters(Pos)) = i; end % 保存舊的聚類中心 OldCenters = Centers; for i = 1:ClusterNum Index = IndexInClusters(i,1:NumberInClusters(i)); Centers(:,i) = mean(SamIn(:,Index)')'; end % 判斷新舊聚類中心是否一致，是則結(jié)束聚類 EqualNum = sum(sum(Centers==OldCenters)); if EqualNum == InDim*ClusterNum, break, end end % 計算各隱節(jié)點的擴展常數(shù)（寬度） AllDistances = dist(Centers',Centers); % 計算隱節(jié)點數(shù)據(jù)中心間的距離（矩陣） Maximum = max(max(AllDistances)); % 找出其中最大的一個距離 for i = 1:ClusterNum % 將對角線上的0 替換為較大的值 AllDistances(i,i) = Maximum+1; end Spreads = Overlap*min(AllDistances)'; % 以隱節(jié)點間的最小距離作為擴展常數(shù) % 計算各隱節(jié)點的輸出權(quán)值 Distance = dist(Centers',SamIn); % 計算各樣本輸入離各數(shù)據(jù)中心的距離 SpreadsMat = repmat(Spreads,1,SamNum); HiddenUnitOut = radbas(Distance./SpreadsMat); % 計算隱節(jié)點輸出陣 HiddenUnitOutEx = [HiddenUnitOut' ones(SamNum,1)]'; % 考慮偏移 W2Ex = SamOut*pinv(HiddenUnitOutEx); % 求廣義輸出權(quán)值 W2 = W2Ex(:,1:ClusterNum); % 輸出權(quán)值 B2 = W2Ex(:,ClusterNum+1)