BP神经网络程序问题求助

T-Eric

这是我从书上打下来的代码，因为书的运行软件版本比较低，我的是Malab2013a的版本，所以我修改了部分程序，但运行出来有问题。请哪位高手帮忙指教和修正一下。谢谢！！

1 X: O- t2 |( o! ]%原始数据输入
+ k1 U/ U" V% j0 N# E5 K' fclc
. y6 M2 e3 T7 l6 B4 C' o6 _sqrs=[20.55,22.44,25.73,27.13,29.45,30.1,30.96,34.06,36.42,38.09,39.13,39.99,...
    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
sqjdcs=[0.6,0.75,0.85,0.90,1.05,1.35,1.45,1.60,1.70,1.85,2.15,2.20,2.25,2.35,...
' c5 q+ l: A! \9 S* n    2.5,2.6,2.7,2.85,2.95,3.10];
sqglmj=[0.09,0.11,0.11,0.14,0.20,0.23,0.23,0.32,0.32,0.34,0.36,0.36,0.38,0.49,...
' Y% v* c$ f: `  F* N    0.56,0.59,0.59,0.67,0.69,0.79];
+ a6 \# ?$ h5 N1 lglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
    22598,25107,33442,36836,40548,42927,43462];
5 H  d! m8 C1 J- Hglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    13320,16762,18673,20724,20803,21804];
; z  u" b1 ~- x" [p=[sqrs;sqjdcs;sqglmj];
5 P  b' C# r% G! Ot=[glkyl;glhyl];
! u0 d/ r1 C3 s" b
%数据归一化
, S0 Z- w4 d: m* `[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
dx=[-1,1;-1,1;-1,1];
3 _' ^& D& _3 U! F
%BP网络训练
; g. h. Y5 G( n% Rnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
6 \* t  O  s0 Xnet.trainParam.show=1000;
net.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
5 f! f& @0 b( ?net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
# s- ^) Q& Y) a
%利用原数据对BP网络仿真
3 A' c* u: n( K8 z" y- P/ v& ]an=sim(net,pn);
; R- k8 q) p; g3 |% La=mapminmax('reverse',an,ps2);
3 E1 R% B; c; l" N: s
%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,:);
newh=a(2,:);
/ o, s, ^) w1 w' l  J  E" q# ifigure(2);
5 w; F6 T6 O1 I/ ]1 o: K! Csubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
legend('网络输出客运量','实际客运量');
. j/ J& D) ?5 }  fxlabel('年份');ylabel('货运量、万人');
title('运用工具箱客运量学习和测试对比图');
subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
legend('网络输出货运量','实际货运量');
2 g/ I3 `3 u& T+ l7 @6 J" a; \xlabel('年份');ylabel('货运量、万吨');
6 h5 O" ]. y+ h# w7 k1 ~2 Qtitle('运用工具箱货运量学习和测试对比图');
: n5 S/ v! b+ P; p1 n# W! ~
%新数据仿真
7 S8 N8 `1 D0 @3 X/ upnew=[73.39,75.55
# P5 u3 R' S/ o    3.9635,4.0975
    0.9880,1.0268];
$ z- ^; u% e# rpnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
' q, R( x' U/ d
7 @! ]) a7 j1 u3 I* T1 A

gancm

看了图我感觉是网络输出后面那几个值太大了，前面基本都是一条水平线了，具体哪里还看不出问题。你把修改的地方说一下看看

529084167

T-Eric

gancm 发表于 2014-7-19 12:30
看了图我感觉是网络输出后面那几个值太大了，前面基本都是一条水平线了，具体哪里还看不出问题。你把修改的 ...

因为版本更新，Matlab中的Newff命令用法有所改变，原命令为：
3 K9 @9 U: i% K, }* a: Tnet=newff(dx,[3,7,2]{'tansig','tansig','purelin'},'traingdx');
我修改为：
* ^1 [4 N8 y2 t; Cnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
然后有些命令被其他命令替代了，其中有premnmx，postmnmx，tramnmx命令好像被替换了，使用了一个强大的命令mapminmax。原命令为：
[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
a=postmnmx(an,mint,maxt);
pnewn=tramnmx(pnew,minp,maxp);
8 b- a9 d* c( v% oanew=postmnmx(anewn,mint,maxt)
我修改为：
[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
  a! ^' G4 _& C# x5 ga=mapminmax('reverse',an,ps2);
pnewn=mapminmax('apply',pnew,ps1);
( ]' Z$ Y- Z/ |& E8 zanew=mapminmax('reverse',anewn,ps2)

  [; R+ [+ s( E原程序为：
4 N( J/ i  I& a0 S! f/ o%原始数据输入
clc
sqrs=[20.55,22.44,25.73,27.13,29.45,30.1,30.96,34.06,36.42,38.09,39.13,39.99,...
    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
" a- Y3 t; g: S" R& Osqjdcs=[0.6,0.75,0.85,0.90,1.05,1.35,1.45,1.60,1.70,1.85,2.15,2.20,2.25,2.35,...
    2.5,2.6,2.7,2.85,2.95,3.10];
" [+ C; H. s7 J, o3 W7 J- G; Isqglmj=[0.09,0.11,0.11,0.14,0.20,0.23,0.23,0.32,0.32,0.34,0.36,0.36,0.38,0.49,...
0 Q3 H$ i/ Q' P0 f8 e8 C% N    0.56,0.59,0.59,0.67,0.69,0.79];
1 i. V6 G+ T0 T5 `8 a+ P' Zglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
9 D- J. ^% g7 m    22598,25107,33442,36836,40548,42927,43462];
glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
. `4 k) `' Q% O# f% I    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
9 [- ^0 `4 K* h8 D3 U& }6 s) St=[glkyl;glhyl];
3 {2 g% ?: R4 f4 r7 {6 k
%数据归一化
) s/ t4 U) {6 Z6 v" I[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
dx=[-1,1;-1,1;-1,1];

/ P9 L% l. j) k$ Y%BP网络训练
- z4 Z. [. J0 z& q# N3 }0 V* Nnet=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
0 L: s2 \% m, ]% l* `1 E3 ~net.trainParam.show=1000;
net.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
9 L' {/ j- l+ r9 A9 wnet.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
/ m$ ]# U7 _3 I) m
5 x. m, x, {% k" c7 F) T%利用原数据对BP网络仿真
an=sim(net,pn);
a=postmnmx(an,mint,maxt);

%仿真结果与原数据对比测试
x=1990:2009;
3 L# p- }5 X- N! fnewk=a(1,;
) |, V0 M& g2 Inewh=a(2,;
2 y: N+ Z- ]0 M) @6 m( I( D# ^figure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
* B# A' @) \' ]. h  [legend('网络输出客运量','实际客运量');
# _+ \: v. L! {/ v9 jxlabel('年份');ylabel('货运量、万人');
1 Z3 M0 X) n9 J* R! gtitle('运用工具箱客运量学习和测试对比图');
( ]$ p0 A( O5 G4 K8 ?9 @. f) Usubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
6 ^; h$ {; h, Nlegend('网络输出货运量','实际货运量');
7 L' ~0 L+ W6 Pxlabel('年份');ylabel('货运量、万吨');
/ E4 R& N+ N! ^; q" c" p) l3 Ptitle('运用工具箱货运量学习和测试对比图');

* B1 i/ w) \9 N* v%新数据仿真
pnew=[73.39,75.55
, t$ M/ p) `( }8 x! G& K7 G* Y# E    3.9635,4.0975
    0.9880,1.0268];
pnewn=tramnmx(pnew,minp,maxp);
! }: s0 i$ o& c. n/ W' u0 R5 _0 [1 Panewn=sim(net,pnewn);
anew=postmnmx(anewn,mint,maxt)

修改后程序为：
. y2 t2 N: `9 @- b: o%原始数据输入
$ I3 x% t. R# ], F8 X& k+ ^clc
sqrs=[20.55,22.44,25.73,27.13,29.45,30.1,30.96,34.06,36.42,38.09,39.13,39.99,...
3 Z* e9 q* ~) m0 b" ^: t    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
  g% T% X; J4 l' `sqjdcs=[0.6,0.75,0.85,0.90,1.05,1.35,1.45,1.60,1.70,1.85,2.15,2.20,2.25,2.35,...
    2.5,2.6,2.7,2.85,2.95,3.10];
& i( R& k/ ?# ^$ Ssqglmj=[0.09,0.11,0.11,0.14,0.20,0.23,0.23,0.32,0.32,0.34,0.36,0.36,0.38,0.49,...
    0.56,0.59,0.59,0.67,0.69,0.79];
& R8 U% h) }% C: X, fglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
    22598,25107,33442,36836,40548,42927,43462];
5 {$ b1 H5 H6 v' Z' Lglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
7 Y' W$ ?* F7 J! v% b0 W! Y    13320,16762,18673,20724,20803,21804];
  p! d3 @# b0 a& H- L7 q% P2 b) op=[sqrs;sqjdcs;sqglmj];
! u# o  z3 s. L) q  J: ]0 Et=[glkyl;glhyl];

%数据归一化
! a, r( ]3 F1 {: `[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
, q0 T; p1 W; H, R/ ~
%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
; h( t0 k$ t  G  _net.trainParam.Lr=0.05;
3 g( F/ R1 A4 ]8 a1 Wnet.trainParam.epochs=50000;
  J! c; ?3 h1 t, m) P! P3 wnet.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
  X+ v+ b+ H: {4 n
%利用原数据对BP网络仿真
an=sim(net,pn);
a=mapminmax('reverse',an,ps2);
7 C0 S* |5 w1 A. }; }8 x& W
%仿真结果与原数据对比测试
7 k9 \5 |4 U% Jx=1990:2009;
& X; W6 g5 q9 ^7 i: jnewk=a(1,;
+ R& ]2 f4 P# J2 t0 ~# anewh=a(2,;
6 t9 E+ V- C3 F2 Vfigure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
$ T+ g/ p' f0 u! z) b8 nlegend('网络输出客运量','实际客运量');
xlabel('年份');ylabel('货运量、万人');
* Y, ]1 e  N0 a7 E8 r$ Ctitle('运用工具箱客运量学习和测试对比图');
1 p5 d0 o+ b4 w' g9 s- Fsubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
legend('网络输出货运量','实际货运量');
" w. p: j9 p- ^$ Cxlabel('年份');ylabel('货运量、万吨');
title('运用工具箱货运量学习和测试对比图');

%新数据仿真
pnew=[73.39,75.55
    3.9635,4.0975
) ?( P5 z) v( y    0.9880,1.0268];
! I  s6 L  R" Z! _* ppnewn=mapminmax('apply',pnew,ps1);
1 A. `: C* n, z) F: L; Xanewn=sim(net,pnewn);
. s3 H  `8 u7 s. \% ]& Hanew=mapminmax('reverse',anewn,ps2)
7 o1 z, F' d( j4 S- O: T（修改的地方用颜色标记了）
麻烦您帮忙指出其中的问题，万分感谢！

且生

求书名及页码……

且生

1. clc
   1 Z3 Y5 D) J& E7 ~0 ?7 @
2. sqrs=[20.55,22.44,25.73,27.13,29.45,30.1,30.96,34.06,36.42,38.09,39.13,39.99,...
   8 c! c3 p5 G7 {. f. P# ]
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   6 E& o6 {3 N$ ~2 x- P
4. sqjdcs=[0.6,0.75,0.85,0.90,1.05,1.35,1.45,1.60,1.70,1.85,2.15,2.20,2.25,2.35,...
   * |; X% d) V8 P$ W8 ^
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   + C* `. S/ p& W9 N' g+ w& m* O
6. sqglmj=[0.09,0.11,0.11,0.14,0.20,0.23,0.23,0.32,0.32,0.34,0.36,0.36,0.38,0.49,...
   ' Q5 ?3 D1 K/ b
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   $ T4 M2 r: ^% w* i5 `* a
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...+ _- j* P4 M7 B2 u) Q& G* ?
   
9.      22598,25107,33442,36836,40548,42927,43462];
   \" u5 s8 F% q% g, d
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    ) L1 I5 d$ F  y
11.      13320,16762,18673,20724,20803,21804];
    8 k1 y  l) u\" ?( l2 n- J- G
12. p=[sqrs;sqjdcs;sqglmj];
    / c; V, \$ ~. v$ B9 l- M) r- U2 w
13. t=[glkyl;glhyl];
    6 ^, M: ?8 E' ^% Y8 q3 k
14. %数据归一化  M+ Z, q$ g0 n& {) ^
    
15. [pn,ps1]=mapminmax(p);
    % l+ _/ l6 a! z2 J5 ~, ^
16. [tn,ps2]=mapminmax(t);
    8 V/ M* w) ]5 I6 T& J1 s# c
17. %dx=[-1,1;-1,1;-1,1];( p) e# C2 D( T* @+ `, O! g
    
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]
    ! \- y  g3 C7 e' ~
19. 2 V$ O/ C' p+ b7 z! \9 B: z2 b
    
20. %BP网络训练  h. `- g8 \# w+ {  f! X( l* E
    
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]
    ' b' }0 n, r\" t. z) Z. i
22. net.trainParam.show=1000;
    ' h$ H& D0 ?0 Z/ W
23. net.trainParam.Lr=0.05;5 C! H. a6 k1 o$ i( W
    
24. net.trainParam.epochs=50000;
    . W  N4 h3 g1 p\" N1 v
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]
    0 O- o' Q4 n4 Q7 Z0 I% J
26. net=train(net,pn,tn);2 c* l: J7 F4 Q6 L& ?( K
    
27. & e$ T+ U8 R1 V! z2 G3 m
    
28. %利用原数据对BP网络仿真) I, n0 j& `5 l% J) }
    
29. an=sim(net,pn);  q; r# E7 s. e* E8 {! o$ m
    
30. a=mapminmax('reverse',an,ps2);4 n& I& _' T8 _  j3 G, i3 h3 {( r: i
    
31. : Q! y+ p7 g, Z1 r
    
32. %仿真结果与原数据对比测试
    / U. [\" v1 z: S! w
33. x=1990:2009;
    1 C+ O3 Z: ^8 x) m+ g
34. newk=a(1,:);$ n0 H/ v' u3 O. V0 `# v2 K  k
    
35. newh=a(2,:);
    . S6 _* `8 x/ _
36. figure(2);
    2 w$ `' P4 g0 R) ~* i; L
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    4 J- k5 V; ]' j$ q0 \; m, Q
38. legend('网络输出客运量','实际客运量');3 Y+ v$ S( s$ _: w
    
39. xlabel('年份');ylabel('货运量、万人');
    ' P. P& n- O+ q: j2 f; u
40. title('运用工具箱客运量学习和测试对比图');. h5 U$ P3 G\" _1 a0 O
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    . e9 g4 G  {5 t7 _' @* f5 O( h
42. legend('网络输出货运量','实际货运量');
    0 ?\" E8 {0 ]; `4 R1 d5 }: }( ]6 n, E
43. xlabel('年份');ylabel('货运量、万吨');; e) W5 n0 }' ?6 ^9 k0 Y8 j8 ?
    
44. title('运用工具箱货运量学习和测试对比图');
    & }: g7 n) y) {/ n
45. & T8 d; w5 C- p; e1 K5 |
    
46. %新数据仿真\" y- h4 Q* I. A8 t5 W! s\" j4 [
    
47. pnew=[73.39,75.55- F/ K9 p# f2 R1 A
    
48.      3.9635,4.0975
    5 z# }: g' B- E1 M: @, J& T0 ~  L+ @
49.      0.9880,1.0268];
    0 Y  i9 X, X* f+ P# n. w
50. pnewn=mapminmax('apply',pnew,ps1);
    * p' j$ D# R0 a' ~( ?0 I9 C& m- W6 a
51. anewn=sim(net,pnewn);
    $ _* T( T  L\" \6 P8 C8 c- l
52. anew=mapminmax('reverse',anewn,ps2)

复制代码

且生

忽略我上面那个，改动的地方在18，21，25行……求批

1. clc
   / C, W% C$ v( Q- `& E7 ?
2. sqrs=[20.55,22.44,25.73,27.13,29.45,30.1,30.96,34.06,36.42,38.09,39.13,39.99,...( c7 J\" a; p, j! }, m
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];7 s9 F/ o* |& {) u
   
4. sqjdcs=[0.6,0.75,0.85,0.90,1.05,1.35,1.45,1.60,1.70,1.85,2.15,2.20,2.25,2.35,...
   * y' U3 h* E/ L) g- O
5.     2.5,2.6,2.7,2.85,2.95,3.10];5 K6 N8 d* V* {2 M
   
6. sqglmj=[0.09,0.11,0.11,0.14,0.20,0.23,0.23,0.32,0.32,0.34,0.36,0.36,0.38,0.49,...
   & ]! U8 j9 X+ [7 C* a3 C; j) G+ L
7.     0.56,0.59,0.59,0.67,0.69,0.79];$ {- z) a8 f1 ]
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,.... Y3 p' o9 A% B3 F9 y+ G6 J# d
   
9.      22598,25107,33442,36836,40548,42927,43462];
   * w5 V9 [1 w7 t; U  j& y, n3 X# l
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    ) m  K# A& Q# X6 @- y1 H
11.      13320,16762,18673,20724,20803,21804];
    1 G9 B' l# n5 C7 B2 M: C+ o
12. p=[sqrs;sqjdcs;sqglmj];4 C0 t. K* N- G
    
13. t=[glkyl;glhyl];: R& t$ R5 ~4 v\" E* h0 N
    
14. %数据归一化, o\" m1 R1 a' E  M& y
    
15. [pn,ps1]=mapminmax(p);
    9 ^7 h; V; a1 j  f% s# ^
16. [tn,ps2]=mapminmax(t);
    3 w: O( A4 o1 N$ v# ]9 N\" G
17. %dx=[-1,1;-1,1;-1,1];% X$ L- c* z; G+ x\" V4 ?
    
18. p0=minmax(pn);t0=minmax(tn);
    9 x% ^4 [/ v1 p9 v
19. 
    ! N& A7 p- o3 J7 N2 C* n, c6 ]
20. %BP网络训练
    7 Z8 \5 P\" W3 h( J( G
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
    9 O3 u8 ~1 n) O
22. net.trainParam.show=1000;8 n, }, g' K2 }5 \& F& X# Y
    
23. net.trainParam.Lr=0.05;& I% l0 h3 ~8 J$ S\" p) z
    
24. net.trainParam.epochs=50000;0 i9 C& _4 D: |7 ?
    
25. net.trainParam.goal=0.65*10^(-5);6 o$ X& v/ e6 p/ F) }/ {
    
26. net=train(net,pn,tn);& r6 v8 T  }8 u& J, M6 a
    
27. 
    6 x# i0 L' I! G! Z5 _
28. %利用原数据对BP网络仿真
    + A- E( r% l* f
29. an=sim(net,pn);
    + n6 Q3 M5 h! _. F. ~
30. a=mapminmax('reverse',an,ps2);4 g5 y6 P3 N% I\" z' m( j
    
31. 
    - s0 V5 t, z! B. `
32. %仿真结果与原数据对比测试 5 {; J' q\" p( \
    
33. x=1990:2009;
    2 v6 t. Z% k4 n4 W/ Y
34. newk=a(1,:);! z; q3 W2 Q0 t! T- N' ~
    
35. newh=a(2,:);
    8 O! t/ S3 Q* P: V/ s1 _8 _0 a# {
36. figure(2);+ f; z: j' C8 ~# Q# I$ D9 i5 Z* s6 p
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');8 O3 j6 {% W: N6 C
    
38. legend('网络输出客运量','实际客运量');
    3 f8 y% Y! i0 [7 K# c
39. xlabel('年份');ylabel('货运量、万人');
      G* h8 S. Y3 C4 B
40. title('运用工具箱客运量学习和测试对比图');- A5 I6 I2 m; G\" K7 e
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    + }4 h2 C$ n& |  b7 P
42. legend('网络输出货运量','实际货运量');
    9 B( a3 j1 p- I
43. xlabel('年份');ylabel('货运量、万吨');0 Z  Z! H8 }% Z& l/ t$ C
    
44. title('运用工具箱货运量学习和测试对比图');
    / m3 d1 q: B, [5 l6 p7 G+ G& D  J  W
45. 7 S: i\" K\" U: h# f! U
    
46. %新数据仿真
    & o8 z- i- n. k  H, r0 j: L: K9 B
47. pnew=[73.39,75.55
    5 H' }# r( C8 r, f, F) o4 J
48.      3.9635,4.0975: N8 L5 y2 x+ f0 N  d( f+ c
    
49.      0.9880,1.0268];
    3 n* p\" I8 ?$ e, z' z
50. pnewn=mapminmax('apply',pnew,ps1);
    9 \! ^; O0 A: V# i: L& a
51. anewn=sim(net,pnewn);
    . @* d\" n2 I7 z6 `& I% n! N3 A0 B& K
52. anew=mapminmax('reverse',anewn,ps2)

复制代码

T-Eric

且生 发表于 2014-7-21 18:22
忽略我上面那个，改动的地方在18，21，25行……求批

确实改善了很多，很是感谢。但效果还是不太理想，而且为何只学习了几十次就停了呢？即使我修改了目标精度。

且生

1. %BP网络训练0 g; j/ \4 f8 X\" c\" A& j
   
2. net=newff(p0,t0,6,{'tansig'},'traingd');
   \" q* J: F; ]$ Q6 N1 s1 K4 m
3. net.trainParam.show=1000;
   , T& K\" e* V' Y( @( x8 o! ?( D
4. net.trainParam.Lr=0.05;
   0 f) a6 ?$ ^; w. S# d2 m
5. net.trainParam.epochs=2000;
   / _. j\" S7 i8 k9 |- N5 M
6. net.trainParam.goal=0.65*10^(-4);
   ' w- o( t4 Y8 e
7. net=train(net,pn,tn);

复制代码

我只改了这里面的，
/ V, R, k" @3 W* D1 E一、改成单隐含层的，6个节点
二、训练函数改成梯度下降BP算法 traingd
三、迭代次数改成2000
上面的参数是自己试的，我也不知道为什么。
# X1 r' Q+ t) [2 [  Q; Q由于这玩意儿比较不靠谱，楼主多运行几次就能找到拟合的比较好的网络。
关键问题是有没有过拟合我也不知道，等大神来解答吧

且生

因为样本比较少，所以训练函数没有选会调节学习率的