神经网络工具箱的使用

天零ヮ冰釋

1. clear
   : W1 S) U3 |3 {
2. load('trainset.mat');                   %加载训练集数据
   3 u3 K* Q! `% O! N
3. load('testset.mat');                    %加载测试集数据8 ^' m2 w/ {' J* d! n3 _\" p
   
4. p1=trainsetdata([1:179],:);% ~7 F: {! [3 P9 R4 H! a
   
5. p2=trainsetdata([180:300],:);) E% i; m3 d+ ?4 D! H
   
6. p=[p1;p2]';\" b4 ~. b* ^- b$ _6 X7 \1 t8 K
   
7. pr=minmax(p);
   ! o9 x' a% M1 Q, v
8. goal=[ones(1,179),zeros(1,121);zeros(1,179),ones(1,121)];
   - i1 x% H& c* m) o
9. net=newff(pr,[21,160],{'tansig','logsig'}); %隐含层节点和输出层节点3 d$ C+ ~0 y$ M: O* D
   
10.                                             %取值不太确定，请指教
    8 ]5 ?5 e5 a1 {5 U7 }: Z
11. net.trainParam.show = 10;) y  {' h& @: S7 C: V9 |
    
12. net.trainParam.lr = 0.05;
    ! a% V4 n, l4 P  B& d) S5 [
13. net.trainParam.goal = 1e-10;
    1 J5 s; t  _. I# J' J) L2 o$ D8 }' `
14. net.trainParam.epochs = 50000;
    ; y3 O  I3 Y3 O& ~
15. net = train(net,p,goal);
    8 ?\" w# |' [+ S( F1 O
16. x=testsetdata';4 _. l0 m' W  l
    
17. y0=sim(net,p)

复制代码

下面是运行结果，请各位高手帮忙改改啊，主要是节点数那块儿不知道该怎么取值

2. Warning: NEWFF used in an obsolete way.
   
3. ' O$ d\\" l. w! Q' q7 U' u
4. > In obs_use at 18
   
5. $ L4 [& O5 z% C4 x) V& b+ E/ ?
6.   In newff>create_network at 127
   
7. 5 I9 a7 C4 J# N1 \
8.   In newff at 102: Y7 v0 p  z5 K- f# [* T, n! E. j! i
   
10.   In test at 9: k6 w. r+ d$ \  D
    
12.           See help for NEWFF to update calls to the new argument list.
    
13. 3 v4 S* l* S( ~1 f' c$ h
14. - ]3 H  o# _$ |# {0 G; a- I
    
16. ??? Error using ==> trainlm at 109
    
17. ; A$ M0 e& f  Z
18. Output data size does not match net.outputs{2}.size.
    
19. # o. }) N& q# E8 ]% S& ?
20. 
    
21.   u( H  J  c  g) Y- K1 p
22. Error in ==> network.train at 1075 }+ b\\" S4 `; h) ~+ f3 d8 j
    
24. [net,tr] = feval(net.trainFcn,net,X,T,Xi,Ai,EW,net.trainParam);
    
25. + g/ D/ ^! F\\" g% r  n
26. 
    
27. 2 j\\" M+ D2 ~( A: ]6 F\\" \/ U7 S
28. Error in ==> test at 15# I# t3 c( P! ?& U
    
30. net = train(net,p,goal);

Warning: NEWFF used in an obsolete way. <br /> <span style=\\"display:none\\">' O$ d\\" l. w! Q' q7 U' u</span> > In obs_use at 18<br /> <span style=\\"display:none\\">$ L4 [& O5 z% C4 x) V& b+ E/ ?</span>   In newff>create_network at 127<br /> <span style=\\"display:none\\">5 I9 a7 C4 J# N1 \</span>   In newff at 102<font class=\\"jammer\\">: Y7 v0 p  z5 K- f# [* T, n! E. j! i</font><br />   In test at 9<font class=\\"jammer\\">: k6 w. r+ d$ \  D</font><br />           See help for NEWFF to update calls to the new argument list.<br /> <span style=\\"display:none\\">3 v4 S* l* S( ~1 f' c$ h</span> <font class=\\"jammer\\">- ]3 H  o# _$ |# {0 G; a- I</font><br /> ??? Error using ==> trainlm at 109<br /> <span style=\\"display:none\\">; A$ M0 e& f  Z</span> Output data size does not match net.outputs{2}.size.<br /> <span style=\\"display:none\\"># o. }) N& q# E8 ]% S& ?</span> <br /> <span style=\\"display:none\\">  u( H  J  c  g) Y- K1 p</span> Error in ==> network.train at 107<font class=\\"jammer\\">5 }+ b\\" S4 `; h) ~+ f3 d8 j</font><br /> [net,tr] = feval(net.trainFcn,net,X,T,Xi,Ai,EW,net.trainParam);<br /> <span style=\\"display:none\\">+ g/ D/ ^! F\\" g% r  n</span> <br /> <span style=\\"display:none\\">2 j\\" M+ D2 ~( A: ]6 F\\" \/ U7 S</span> Error in ==> test at 15<font class=\\"jammer\\"># I# t3 c( P! ?& U</font><br /> net = train(net,p,goal);

jiajinshan110

..................