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！！
$ M% N* y% t4 {# C
4 O0 F+ a, i  M3 S9 {. v& Q: A%原始数据输入
9 d" ]+ B4 L9 a& K6 u& Wclc
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,...
: D/ [$ V% M# d2 D  B+ |    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
9 g$ m# S3 |" nsqjdcs=[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];
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,...
+ ^8 @# F9 Q+ f0 ^6 V9 \    0.56,0.59,0.59,0.67,0.69,0.79];
glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
. {6 A4 C: F3 x! D3 U' S3 [- P5 x    22598,25107,33442,36836,40548,42927,43462];
glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
' Y7 `! n, l7 h    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
% D' `3 M+ y: r. M# Q4 ]t=[glkyl;glhyl];
: C0 I1 \& e1 Z  L- V1 V
%数据归一化
& c8 w! m, M2 M# r[pn,ps1]=mapminmax(p);
4 f+ |9 ~" Q; M% `. t  Y$ o[tn,ps2]=mapminmax(t);
dx=[-1,1;-1,1;-1,1];
( @9 U. {. G& w* Y; y$ ^
%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
2 p4 x% ]+ L5 g# Z* [# f3 {% Pnet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
* W: Q% T# p! _  Rnet.trainParam.goal=0.65*10^(-3);
5 X0 f! U( ^  o: t4 inet=train(net,pn,tn);
2 u3 y% \' b% x3 L# ?
%利用原数据对BP网络仿真
, I, P$ q1 ?0 V. P# ran=sim(net,pn);
7 t& x1 Z# l2 ?' h3 Pa=mapminmax('reverse',an,ps2);
$ j0 A/ D! @0 C7 \1 n" k; F- k
%仿真结果与原数据对比测试
x=1990:2009;
/ X( W+ d+ S: B1 c; P1 gnewk=a(1,:);
- h! Q9 r6 P! h/ \+ n0 vnewh=a(2,:);
figure(2);
5 H3 k- H; k6 R4 Psubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
legend('网络输出客运量','实际客运量');
xlabel('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');
xlabel('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');
; E! s+ s9 G$ J/ K$ e) w
%新数据仿真
: e9 v) N2 B3 G- S+ ypnew=[73.39,75.55
7 W1 K6 u- \" R$ F    3.9635,4.0975
    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
/ X, k, b+ ?8 ]' G- ]4 g' Qanewn=sim(net,pnewn);
1 b4 I5 ?8 Q: ]anew=mapminmax('reverse',anewn,ps2)
* W# {3 F9 W6 k1 p5 G7 G
2 Z  P. x* e) x4 R

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 ...

, T% Z" J* D5 Q, W6 S2 A; t4 Z! [因为版本更新，Matlab中的Newff命令用法有所改变，原命令为：
5 k& e1 e' D5 cnet=newff(dx,[3,7,2]{'tansig','tansig','purelin'},'traingdx');
我修改为：
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
; S" H" M; U8 n# x) c7 U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：
( m  ~7 e( Y" [# q  n[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
: f& o" f* s' R; _a=postmnmx(an,mint,maxt);
7 x/ ]  Z7 q* C# g1 O* apnewn=tramnmx(pnew,minp,maxp);
anew=postmnmx(anewn,mint,maxt)
6 j) \3 U3 c7 E% l' O我修改为：
" F4 s+ W; X* w, Z1 z5 k[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
0 X( N, H! e: p/ ~3 `% ga=mapminmax('reverse',an,ps2);
0 D6 S# v. Z3 X5 _pnewn=mapminmax('apply',pnew,ps1);
$ l& s/ o# s/ h+ s* h# q0 b& X; kanew=mapminmax('reverse',anewn,ps2)
! v4 e$ z; ?0 W' Q. s, @: x
原程序为：
# T/ w$ H- S" X# m8 ?( w" E%原始数据输入
% ~0 w: ~" I6 |7 _clc
$ d1 w, c' c* z# t. Hsqrs=[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];
% v% O2 W( j6 w+ {8 C0 n; ?1 asqjdcs=[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,...
, u: P( k' D3 U; g5 ?0 W9 |    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,...
    0.56,0.59,0.59,0.67,0.69,0.79];
glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
    22598,25107,33442,36836,40548,42927,43462];
glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

%数据归一化
; w0 `* L8 L& {* b5 {: F[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
dx=[-1,1;-1,1;-1,1];

%BP网络训练
net=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
$ `/ t+ Z* J0 X2 h  Wnet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
* P! Z. }) ?: Rnet.trainParam.goal=0.65*10^(-3);
. L/ m: j7 [! N  q4 y- ynet=train(net,pn,tn);
) l8 K3 h  |1 U& u, q- f8 ^# a, [
% k3 i6 }  l" P2 e2 u7 o$ F5 {%利用原数据对BP网络仿真
5 l+ l4 k  s$ Y8 i9 h5 Kan=sim(net,pn);
a=postmnmx(an,mint,maxt);

( V7 K3 s& x1 P% s  d; W2 n0 r%仿真结果与原数据对比测试
3 Y( J! F8 n# o% @. Bx=1990:2009;
newk=a(1,;
5 e/ P& x1 V) znewh=a(2,;
figure(2);
( d0 n( m% u  h8 {" T, b8 Usubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
( h, Z, Z* n4 q+ f* U# h* Qlegend('网络输出客运量','实际客运量');
, k6 Z0 q0 z8 ]# e8 Qxlabel('年份');ylabel('货运量、万人');
& @0 f2 V7 m/ h1 otitle('运用工具箱客运量学习和测试对比图');
( e/ Q/ A, A! h+ Osubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
" [3 S8 J5 k! N5 R9 W: Qlegend('网络输出货运量','实际货运量');
xlabel('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
( `8 c! }/ l. g    3.9635,4.0975
    0.9880,1.0268];
pnewn=tramnmx(pnew,minp,maxp);
anewn=sim(net,pnewn);
anew=postmnmx(anewn,mint,maxt)
) j' m" S5 q' c+ t% N" W2 O
修改后程序为：
& D8 |1 |3 {7 W/ \' X. A- p%原始数据输入
% `8 Y0 F2 L7 I% j$ fclc
% B4 }" Z9 ?$ E6 h- ~5 [' Tsqrs=[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];
; A5 c. K5 }& x5 M1 Rsqjdcs=[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];
- J2 \1 p* J( B. Z7 Wsqglmj=[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,...
, R# s8 k0 C5 t9 y    0.56,0.59,0.59,0.67,0.69,0.79];
+ ?. \) M) Y* V4 p' F! j/ Bglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
    22598,25107,33442,36836,40548,42927,43462];
glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
' Y3 I4 Z- B1 Q) s  y) k* F    13320,16762,18673,20724,20803,21804];
5 Q2 y. s0 J! `p=[sqrs;sqjdcs;sqglmj];
( [3 y% p* k  et=[glkyl;glhyl];

%数据归一化
[pn,ps1]=mapminmax(p);
$ n# C8 s5 m. L9 S2 o[tn,ps2]=mapminmax(t);
9 h; X' [5 U2 S8 w# F
%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
3 [* }8 q, W# H0 onet.trainParam.Lr=0.05;
/ y& |) u& l3 M/ f- nnet.trainParam.epochs=50000;
* @8 H, Y! |3 ?& x  x* \net.trainParam.goal=0.65*10^(-3);
1 z% e# N$ }5 G0 dnet=train(net,pn,tn);
6 `% x% f2 e4 j( K
%利用原数据对BP网络仿真
an=sim(net,pn);
a=mapminmax('reverse',an,ps2);
, S0 N: J6 a+ [* \3 O) R1 X
%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,;
4 l) Q5 \/ w, r5 e4 qnewh=a(2,;
figure(2);
$ V5 ~% A4 g2 I; D# nsubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
, g" H( ?+ k& k4 N' P+ J- olegend('网络输出客运量','实际客运量');
xlabel('年份');ylabel('货运量、万人');
3 y: N# n; i& C1 B" Y# x8 qtitle('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--+');
% Q& I4 I( g3 K$ v. d' Alegend('网络输出货运量','实际货运量');
xlabel('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');
( p. m* V4 X; ?' h8 |& d) M
# D" A# o' \" V: V( a%新数据仿真
pnew=[73.39,75.55
9 G" I1 f5 E* y0 U4 x  n  x    3.9635,4.0975
    0.9880,1.0268];
/ s  l$ T+ ?$ v9 H  J) x9 dpnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
3 h0 _2 ^& L. e2 W* G. j（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
   * @7 E6 M& l0 U( ~! T\" n5 i2 R
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,...& ~5 l' n4 y# x/ F
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   % \# @, i  I6 a
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,...
   & Z- h7 w! E/ I7 j+ W% i\" e% c
5.     2.5,2.6,2.7,2.85,2.95,3.10];\" m  K9 l& T( i5 e
   
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,...( T' L  _0 k, D& q\" d. R
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   3 r- z! a, [! B3 H4 B( v
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   ! Y1 F\" M9 n! K- v' h  Q
9.      22598,25107,33442,36836,40548,42927,43462];
   & F, }* m; J  v  O
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    / A5 m7 G* w9 x, U# e) B+ y
11.      13320,16762,18673,20724,20803,21804];
    . W1 t: P' i- x
12. p=[sqrs;sqjdcs;sqglmj];! ^' [1 {* r6 T( Z4 G6 K, r( s. B3 k# {
    
13. t=[glkyl;glhyl];7 H& h' r; z) v: F  b, f: E
    
14. %数据归一化
    8 T3 D* p( V0 w' k2 j! N
15. [pn,ps1]=mapminmax(p);
    $ o) M! r% c# y; R' q1 B
16. [tn,ps2]=mapminmax(t);
    + a6 z! _+ M1 A* ~* n
17. %dx=[-1,1;-1,1;-1,1];( F6 I2 i, c/ D$ p5 b* m
    
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]
    8 P9 {  C5 m\" E7 {- B9 A\" V
19. % y6 [* a9 a- ?# d\" g. U! j
    
20. %BP网络训练3 U  x( j4 G! a
    
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]
    $ l, t) W0 q0 b5 g0 s3 o
22. net.trainParam.show=1000;
    # z& Y\" ^: w( u  P5 b, s4 t) v
23. net.trainParam.Lr=0.05;
    ; `1 p& v8 |% N- X  I3 E
24. net.trainParam.epochs=50000;
    + R, ~# _5 ]( f& G
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]
    \" q: ?6 `. a\" S& q& F\" O, D4 L( E
26. net=train(net,pn,tn);
    3 V! f0 J: Y- Z' u: |' y& S
27. 
    # c' E0 _( p. ~6 d2 y* y
28. %利用原数据对BP网络仿真
    $ D) J4 ?; M; v\" \
29. an=sim(net,pn);
      P( D4 s2 a  }\" N- G
30. a=mapminmax('reverse',an,ps2);
    # u  \5 q- g! e
31. 
    / t+ ]7 I7 v3 h5 n# p4 g, U' |# w
32. %仿真结果与原数据对比测试 % @4 T  w) O- n9 j
    
33. x=1990:2009;9 i6 K3 P0 [% O9 P( `8 u
    
34. newk=a(1,:);
    8 \\" H, F5 p) \' `, C
35. newh=a(2,:);9 a+ V/ N$ W  F+ H5 G/ ]
    
36. figure(2);. y$ ^6 o+ W8 {2 d! x
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');3 r/ A2 m9 j: ]\" Z  s+ D3 t3 L
    
38. legend('网络输出客运量','实际客运量');
    ) g9 E2 h# q0 j5 B* {7 \2 R1 d
39. xlabel('年份');ylabel('货运量、万人');
    * Q: @9 t( U8 ~7 g8 Y+ [
40. title('运用工具箱客运量学习和测试对比图');; A2 L; z, v6 K% ~' J; g
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    : V/ T$ [0 {! P, _
42. legend('网络输出货运量','实际货运量');( X6 }$ [. e+ p5 y( I
    
43. xlabel('年份');ylabel('货运量、万吨');
    2 B2 @/ ?3 D  i) I
44. title('运用工具箱货运量学习和测试对比图');6 D+ A# ^' H2 a& u0 v! W8 B3 I
    
45. : p+ u# Y% M: Z( ]# R- F6 {, X7 ~% D
    
46. %新数据仿真
    5 M$ g, |) C0 L# u
47. pnew=[73.39,75.55
    2 a1 J0 y( S\" v3 M, Q) w! @; U; _+ P
48.      3.9635,4.0975
    ' T0 S& A: u: \) o: e: S3 @4 g. x  {
49.      0.9880,1.0268];$ V& U1 l, ^7 W% }
    
50. pnewn=mapminmax('apply',pnew,ps1);, w$ j# m+ Q2 x+ F7 C, _\" j$ h: M/ i
    
51. anewn=sim(net,pnewn);4 ?  d5 n/ B# [' P2 w
    
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
   \" x* D! O' ~6 ~8 [/ O3 W$ z7 ^% ~
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,...: K6 t$ K9 S# P2 s6 m/ K4 y
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];0 E9 ?5 g; Z& I
   
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,...7 k+ J: M' B9 [8 [; g7 D* m
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   # g- T3 `3 C( u  D0 r3 D2 p9 C
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,...9 F) [  _3 i& j  e' n8 D
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   & V! j8 Y8 H& v* ]3 C
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   / _' v' r- D! p) U
9.      22598,25107,33442,36836,40548,42927,43462];( t3 t+ N( O. }1 V
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    ! {- B' @\" Z, N& K- E. K
11.      13320,16762,18673,20724,20803,21804];
    % x0 L6 j# j+ F) s+ \3 N; `6 O
12. p=[sqrs;sqjdcs;sqglmj];8 T, |+ d& t+ N4 }+ r
    
13. t=[glkyl;glhyl];
    ) u& U* Q, n( l/ H: |
14. %数据归一化
    1 x5 I( u# Y( Z0 f) ]- u2 E
15. [pn,ps1]=mapminmax(p);3 C  Z) v' @4 p3 d, ]& W
    
16. [tn,ps2]=mapminmax(t);+ ^5 _% p- e$ g$ w- U2 U
    
17. %dx=[-1,1;-1,1;-1,1];
    + d$ N) m\" b9 D8 n
18. p0=minmax(pn);t0=minmax(tn);
    & R  J& G) t+ t0 v
19.   x& q% ^$ z! ?4 a
    
20. %BP网络训练
    5 O9 g# D  Z. @0 t; H# B
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); . L, y5 N$ k( L
    
22. net.trainParam.show=1000;
    $ b; O( z2 h# c, [- a, o
23. net.trainParam.Lr=0.05;
    $ g3 I# d7 z; a9 b
24. net.trainParam.epochs=50000;+ D\" q3 ]\" t# b* Y, u
    
25. net.trainParam.goal=0.65*10^(-5);% j; U/ u2 U, y& D\" S. v4 t% i$ u0 c
    
26. net=train(net,pn,tn);
    3 y1 I; f& [' M) u: ?2 F2 F' a
27. 
    $ K  T% l9 e; p8 u9 p! y
28. %利用原数据对BP网络仿真\" T6 n( i* v% W2 O% r
    
29. an=sim(net,pn);
    % q* G' i$ P; w* s$ x4 I/ C! L
30. a=mapminmax('reverse',an,ps2);
    / Y' `3 g7 f, f3 C- M8 u5 y
31. 
    \" f5 S+ T. D) m$ Y& t) Y
32. %仿真结果与原数据对比测试
    , l7 ?5 N! b! n
33. x=1990:2009;, C$ a% L  _* o; t1 Y: |+ z
    
34. newk=a(1,:);
    $ M) @5 r' O, a
35. newh=a(2,:);
    2 [1 Z  R/ q% R- q) [. g  B, r; \
36. figure(2);9 e\" e+ ]: _( r\" I& _$ K8 A
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    : p# L1 z2 R9 B& [1 [
38. legend('网络输出客运量','实际客运量');( L9 p) c1 C; ?. k# D
    
39. xlabel('年份');ylabel('货运量、万人');
      l% f  V2 C5 m6 ^- o
40. title('运用工具箱客运量学习和测试对比图');
    5 V4 E5 i6 }) P
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    ( T: D' U! X0 D8 Y5 k+ a* C
42. legend('网络输出货运量','实际货运量');$ q$ f& P! v3 D- G$ P7 @1 _
    
43. xlabel('年份');ylabel('货运量、万吨');
    % v7 d/ `5 n( Y1 Q
44. title('运用工具箱货运量学习和测试对比图');
    \" r\" X8 d# Q* x$ ?6 x
45. 
    8 ?' M3 {9 w! Z
46. %新数据仿真0 [( Q+ }9 t! b' c( l
    
47. pnew=[73.39,75.55! Q; B9 u  v1 R$ S2 N2 |& t+ E
    
48.      3.9635,4.0975
    * Q. Q4 q2 q! x! N
49.      0.9880,1.0268];9 I, j6 _3 Y6 p$ w3 v
    
50. pnewn=mapminmax('apply',pnew,ps1);
    ' E$ \) Y3 h. P! `
51. anewn=sim(net,pnewn);
    & _; p  u' W7 h
52. anew=mapminmax('reverse',anewn,ps2)

复制代码

T-Eric

且生 发表于 2014-7-21 18:22
4 z' e& C* T* @/ H# a6 j  u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
   : t5 [+ [+ d/ g1 k8 g5 |\" R/ K# }
2. net=newff(p0,t0,6,{'tansig'},'traingd'); & j+ d0 E) U8 \& K( r
   
3. net.trainParam.show=1000;/ x6 |0 E! w7 c5 a  t, f, M\" ^
   
4. net.trainParam.Lr=0.05;8 \5 `7 ?: `& l) m4 r: m. I
   
5. net.trainParam.epochs=2000;
   / s\" M9 |( W' q$ ]4 L# H
6. net.trainParam.goal=0.65*10^(-4);' ^; Y# A8 F# C1 ]1 o
   
7. net=train(net,pn,tn);

复制代码

我只改了这里面的，
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
, @, S7 |0 F  Q+ r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。
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