新版Matlab中神經(jīng)網(wǎng)絡(luò)訓(xùn)練函數(shù)Newff的使用方法

1、新版Matlab中神經(jīng)網(wǎng)絡(luò)訓(xùn)練函數(shù)Newff的使用方法一、 介紹新版newffSyntax· net = newff(P,T,S1 S2.S(N-l),TF1 TF2.TFNl, BTF,BLF,PF,IPF,OPF,DDF)Descriptionnewff(P,T,S1 S2.S(N-l),TF1 TF2.TFNl, BTF,BLF,PF,IPF,OPF,DDF) takes several argumentsPR x Q1 matrix of Q1 sample R-element input vectorsTSN x Q2 matrix of Q2 sample SN-elem

2、ent target vectorsSiSize of ith layer, for N-1 layers, default = .(Output layer size SN is determined from T.)TFiTransfer function of ith layer. (Default = 'tansig' forhidden layers and 'purelin' for output layer.)BTFBackpropagation network training function (default = 'trainlm&#

3、39;)BLFBackpropagation weight/bias learning function (default = 'learngdm')IPFRow cell array of input processing functions. (Default = 'fixunknowns','removeconstantrows','mapminmax')OPFRow cell array of output processing functions. (Default = 'removeconstantrows&#

4、39;,'mapminmax')DDFData divison function (default = 'dividerand')ExamplesHere is a problem consisting of inputs P and targets T to be solved with a network.· P = 0 1 2 3 4 5 6 7 8 9 10;T = 0 1 2 3 4 3 2 1 2 3 4;Here a network is created with one hidden layer of five neurons.

5、3; net = newff(P,T,5);The network is simulated and its output plotted against the targets.· Y = sim(net,P);plot(P,T,P,Y,'o')The network is trained for 50 epochs. Again the network's output is plotted.· net.trainParam.epochs = 50;net = train(net,P,T);Y = sim(net,P);plot(P,T,P,Y,

6、'o')二、 新版newff與舊版newff調(diào)用語法對比Example1比如輸入input（6*1000），輸出output為（4*1000），那么舊版定義：net=newff(minmax(input),14,4,'tansig','purelin','trainlm');新版定義：net=newff(input,output,14,'tansig','purelin','trainlm');Example2比如輸入input（6*1000），輸出output為（4*1000），那

7、么舊版定義：net=newff(minmax(input),49,14,4,'tansig','tansig','tansig','traingdx');新版定義：net=newff(input,output, 49,14, 'tansig','tansig','tansig','traingdx');三、 舊版newff使用方法在新版本中使用提示：舊版本定義的newff雖也能在新版本中使用，但會有警告，警告如下：Warning: NEWFF used in an o

8、bsolete way. > In obs_use at 18  In newff>create_network at 127  In newff at 102          See help for NEWFF to update calls to the new argument list.四、 新版newff與舊版newff使用的訓(xùn)練效果對比舊版本：舊用法訓(xùn)練次數(shù)多，但精度高 新版本：新用法訓(xùn)練次數(shù)少，但精度可能達(dá)不到要求造成上述原因是：程序里面的權(quán)值、閾值的初始值是隨機(jī)賦

9、值的，所以每次運(yùn)行的結(jié)果都會不一樣，有好有壞。你可以把預(yù)測效果不錯的網(wǎng)絡(luò)的權(quán)值和閾值作為初始值。具體可以查看net.iw1,1、net.lw2,1、net.b1、net.b2的值?，F(xiàn)在給一個完整的例子 % 清空環(huán)境變量clcclear% 訓(xùn)練數(shù)據(jù)預(yù)測數(shù)據(jù)data=importdata('test.txt');%從1到768間隨機(jī)排序k=rand(1,768);m,n=sort(k);%輸入輸出數(shù)據(jù)input=data(:,1:8);output =data(:,9);%隨機(jī)提取500個樣本為訓(xùn)練樣本，268個樣本為預(yù)測樣本input_train=input(n(1:500),:

10、)'output_train=output(n(1:500),:)'input_test=input(n(501:768),:)'output_test=output(n(501:768),:)'%輸入數(shù)據(jù)歸一化inputn,inputps=mapminmax(input_train);% BP網(wǎng)絡(luò)訓(xùn)練% %初始化網(wǎng)絡(luò)結(jié)構(gòu)net=newff(inputn,output_train,10);net.trainParam.epochs=1000;net.trainParam.lr=0.1;net.trainParam.goal=0.0000004;% 網(wǎng)絡(luò)訓(xùn)練net

11、=train(net,inputn,output_train);% BP網(wǎng)絡(luò)預(yù)測%預(yù)測數(shù)據(jù)歸一化inputn_test=mapminmax('apply',input_test,inputps);%網(wǎng)絡(luò)預(yù)測輸出BPoutput=sim(net,inputn_test);% 結(jié)果分析%根據(jù)網(wǎng)絡(luò)輸出找出數(shù)據(jù)屬于哪類BPoutput(find(BPoutput<0.5)=0;BPoutput(find(BPoutput>=0.5)=1;% 結(jié)果分析%畫出預(yù)測種類和實際種類的分類圖figure(1)plot(BPoutput,'og')hold onplot(output_test,'r*');legend('預(yù)測類別','輸出類別')title('BP網(wǎng)絡(luò)預(yù)測分類與實際類別比對','fontsize',12)ylabel('類別標(biāo)簽','fontsize',12)xlabel('樣本數(shù)目','fontsize',12)ylim(-0.5 1.5)%預(yù)測正確率rightnum