bp_神經(jīng)網(wǎng)絡(luò)的設(shè)計實例(matlab編程)

1、BP神經(jīng)網(wǎng)絡(luò)的設(shè)計實例神經(jīng)網(wǎng)絡(luò)的設(shè)計實例(MATLAB編程編程)例1采用動量梯度下降算法訓(xùn)練BP網(wǎng)絡(luò)。訓(xùn)練樣本定義如下：輸入矢量為p=[12311153]目標矢量為t=[1111]解：本例的MATLAB程序如下：closeallclearechoonclc%NEWFF——生成一個新的前向神經(jīng)網(wǎng)絡(luò)%TRAIN——對BP神經(jīng)網(wǎng)絡(luò)進行訓(xùn)練%SIM——對BP神經(jīng)網(wǎng)絡(luò)進行仿真pause%敲任意鍵開始clc%定義訓(xùn)練樣本%P為輸入矢量P=[1231

2、1153]%T為目標矢量T=[1111]pauseclc%創(chuàng)建一個新的前向神經(jīng)網(wǎng)絡(luò)=newff(minmax(P)[31]tansigpurelintraingdm)%當前輸入層權(quán)值和閾值inputWeights=.IW11inputbias=.b1%當前網(wǎng)絡(luò)層權(quán)值和閾值layerWeights=.LW21layerbias=.b2pauseclc%設(shè)置訓(xùn)練參數(shù).trainParam.show=50.trainParam.lr=0.05

3、.trainParam.mc=0.9.trainParam.epochs=1000.trainParam.goal=1e3pauseclc%調(diào)用TRAINGDM算法訓(xùn)練BP網(wǎng)絡(luò)[tr]=train(PT)pause=newff(minmax(P)[201]tansigpurelin)pauseclcechooffclcdisp(1.LM優(yōu)化算法TRAINLM)disp(2.貝葉斯正則化算法TRAINBR)choice=input(請選擇

4、訓(xùn)練算法(12):)figure(gcf)if(choice==1)echoonclc%采用LM優(yōu)化算法TRAINLM.trainFcn=trainlmpauseclc%設(shè)置訓(xùn)練參數(shù).trainParam.epochs=500.trainParam.goal=1e6=init()%重新初始化pauseclcelseif(choice==2)echoonclc%采用貝葉斯正則化算法TRAINBR.trainFcn=trainbrpause