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！！
3 `: i3 d! n2 r
& N" [& S; I* u- F( W%原始数据输入
% W  T8 j) ]" e5 ^- nclc
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,...
4 h6 z0 J3 X; k- V) T5 B7 n    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
% u6 V3 e, `" z7 `% E" F$ {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];
4 t, R5 Y% G; I6 `* w# @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,...
5 F' y& Z7 a& I1 V! ?; }! Q. h    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,...
; W! C/ A! k# C& D9 M6 T8 [    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

; \# e% o( _3 W% S%数据归一化
[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
( p  o3 k. F7 h; o4 {' }7 s: ydx=[-1,1;-1,1;-1,1];
+ J8 k' j+ {: S3 Q7 G' U( U
%BP网络训练
% ?2 L3 [' S# _% u# Nnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
4 c; Y( P1 v# q7 r& x( pnet.trainParam.Lr=0.05;
' |# r# }8 |+ Y0 d! bnet.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
; g# W9 G, v1 O. I: H$ H5 I* v& jnet=train(net,pn,tn);
) C- U) P2 _& L- T
%利用原数据对BP网络仿真
& u- t) B* D' l; ?1 Yan=sim(net,pn);
a=mapminmax('reverse',an,ps2);

# q1 M2 l8 i+ ?6 W  W# G8 N; b%仿真结果与原数据对比测试
; S; C: n: t6 u4 N0 w  Fx=1990:2009;
newk=a(1,:);
newh=a(2,:);
figure(2);
subplot(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--+');
  {# F- b, ]1 f+ C9 e, Q# {0 Elegend('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');

7 O+ i/ x0 l2 |- X' j- J2 p%新数据仿真
pnew=[73.39,75.55
  C4 W0 f# R9 c0 w    3.9635,4.0975
6 \( E! E: r% t& G7 d    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
! }  Z; q" R1 v5 ~1 \* V! wanew=mapminmax('reverse',anewn,ps2)
4 r) v" o2 l! E& P2 ~" m8 p

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：
net=newff(dx,[3,7,2]{'tansig','tansig','purelin'},'traingdx');
我修改为：
' ]  n1 d+ E8 J0 i7 z7 o- unet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
$ x" n  I7 K0 }5 k0 q1 T) }  A7 r" Q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);
anew=postmnmx(anewn,mint,maxt)
% \* @# u# K8 c' x# F我修改为：
[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
, w% U+ E" g- P/ J. o0 Z, za=mapminmax('reverse',an,ps2);
pnewn=mapminmax('apply',pnew,ps1);
8 I9 X- b( v$ Kanew=mapminmax('reverse',anewn,ps2)
& |. i# V& O1 q5 K" Q1 }
3 U0 ^. v3 s4 X* ~0 B3 H0 o( T$ ]原程序为：
%原始数据输入
( \$ Y9 G* d9 ^) b( y: ^: Pclc
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,...
( c4 L* m' D; d& d$ f) H" v8 {    2.5,2.6,2.7,2.85,2.95,3.10];
; J# Z3 X* s+ fsqglmj=[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];
0 j! x' Q0 }9 Z5 b% Fglkyl=[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];
( c4 r5 o# X# G; M4 w6 a) K- c# Cp=[sqrs;sqjdcs;sqglmj];
. _) D  j. ?( S' V( R3 ^% ]t=[glkyl;glhyl];

5 N% ?. E' n9 r/ F" q%数据归一化
[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
" m$ M/ J4 w$ _0 g( Rdx=[-1,1;-1,1;-1,1];
) E% W$ L3 K$ R. T' }; c  a, k1 E
%BP网络训练
: k* o3 f3 s! ~4 @0 A9 C' `( d) v( ?net=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
. H; t' N$ e4 V+ M: Rnet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
: w5 }  I7 o  c( ^& Y, ^+ E: z
%利用原数据对BP网络仿真
" ?+ a  D7 `( L: r9 _: nan=sim(net,pn);
  {& P" b+ n& l& g8 t. t8 ca=postmnmx(an,mint,maxt);
/ n1 T* H) ~3 M+ ~  Y# @3 t3 [1 d
%仿真结果与原数据对比测试
- T( S# q1 k, n& O6 E) Ox=1990:2009;
1 m! C! }+ w4 Y( u1 t0 l5 znewk=a(1,;
1 a4 y* L8 K) e5 N& u* ^- I8 wnewh=a(2,;
0 h5 p1 z9 h+ Jfigure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
legend('网络输出客运量','实际客运量');
, U; ^7 f* `, R; C* J2 L2 Xxlabel('年份');ylabel('货运量、万人');
; ~) {: q0 t. N+ `% ^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');
+ g; b6 D8 u7 i7 {7 |7 Z! A% b7 \+ j! S
%新数据仿真
pnew=[73.39,75.55
    3.9635,4.0975
$ A/ R( y, O$ j* {    0.9880,1.0268];
) c* ^  X/ k8 Z  z. zpnewn=tramnmx(pnew,minp,maxp);
, a$ u) j, V% S& eanewn=sim(net,pnewn);
4 V( K8 p- n! s# i4 E- Yanew=postmnmx(anewn,mint,maxt)
" B' i; {1 G' v% m+ g& V. k
修改后程序为：
# E5 P; o8 @5 Y, U1 R%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,...
* A8 X' D4 H$ D/ c  e& W    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
6 U5 l8 Z4 J+ ^* k  b, Tsqjdcs=[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,...
3 s' @) W7 v  d  Z4 ?6 l    2.5,2.6,2.7,2.85,2.95,3.10];
' `! K2 K1 V; F1 G$ l* |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,...
$ L# R- [) H3 _* L6 F+ \% W    0.56,0.59,0.59,0.67,0.69,0.79];
  S4 I3 \) O) Kglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
* `# ]* K5 W% u; c6 S) P2 t3 Q7 A    22598,25107,33442,36836,40548,42927,43462];
glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
# L. \1 A' [0 [; z4 w# H2 v    13320,16762,18673,20724,20803,21804];
" S. ~2 {1 K* T2 d. sp=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

$ B% p) `! I* R5 j%数据归一化
[pn,ps1]=mapminmax(p);
1 w+ m9 ^+ W1 m- V9 @[tn,ps2]=mapminmax(t);
* }% m. F& w- p; w8 ^, X
* `, }% i" }. b- W) D%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
net.trainParam.Lr=0.05;
* w2 e8 Q6 E  W8 Pnet.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
5 r2 b+ U. j8 ~2 i+ p7 _net=train(net,pn,tn);

2 @$ w# x3 U/ c%利用原数据对BP网络仿真
- e9 c4 I; O7 x' Q" s. O, s7 man=sim(net,pn);
a=mapminmax('reverse',an,ps2);

# d0 g( P- a- k" Q%仿真结果与原数据对比测试
* N& ^1 q) L" {3 P7 `, B  [0 sx=1990:2009;
newk=a(1,;
+ _; L' x: y3 ~! ]4 f/ @' s4 G4 Dnewh=a(2,;
figure(2);
subplot(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');
! z0 \3 h- F4 Q4 [& ptitle('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--+');
$ ?2 P/ R: S+ l2 glegend('网络输出货运量','实际货运量');
, g* M: o) x& l, Gxlabel('年份');ylabel('货运量、万吨');
3 [7 H3 Y5 j+ p) n3 Rtitle('运用工具箱货运量学习和测试对比图');
- l2 T0 S. p. o7 a3 W' S* b8 L
6 E9 _' M  f) K* s%新数据仿真
2 }  m( G0 }, p0 e6 M6 h, q' C' Ppnew=[73.39,75.55
    3.9635,4.0975
5 ?9 N- N& g# v  ?    0.9880,1.0268];
" E" l/ N3 E$ ppnewn=mapminmax('apply',pnew,ps1);
% e  ^: P+ b7 A: n. \4 Xanewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
（修改的地方用颜色标记了）
) @; p: b! C1 x  }  U: Q9 H8 r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' [! h* t4 h7 `& I( f; [4 ^: M( N
   
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,...3 ]4 c/ `: d2 e
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   \" p1 Q3 l$ C+ u- [$ `8 w0 D- F! L
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,...
   1 m; }: b  A$ v; Q
5.     2.5,2.6,2.7,2.85,2.95,3.10];# I- o7 A. U5 A1 i
   
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,...
   ! E( |; _% O) i7 W
7.     0.56,0.59,0.59,0.67,0.69,0.79];1 j* O2 H0 ^. l( k
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...( ~\" @% D8 R. H3 d
   
9.      22598,25107,33442,36836,40548,42927,43462];
   9 E: }/ A0 n. ]+ B& v7 {
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    ( w\" |0 [' x! ?
11.      13320,16762,18673,20724,20803,21804];
    ' b% }% [+ y4 A6 s* @
12. p=[sqrs;sqjdcs;sqglmj];
    2 k! v; C+ B3 I) j\" H8 w
13. t=[glkyl;glhyl];/ ~; j  H  E$ e\" b4 M9 `( `
    
14. %数据归一化6 C9 D9 S, d\" j4 V
    
15. [pn,ps1]=mapminmax(p);0 }* ^* v9 S' P0 B
    
16. [tn,ps2]=mapminmax(t);6 S1 j% p6 F$ ?  w# h
    
17. %dx=[-1,1;-1,1;-1,1];
    9 n5 I9 A, b1 W
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]; A, P$ E8 u# M8 V# K
    
19. 
    ) l5 h8 p9 E, x7 p0 \) B, }  F+ w
20. %BP网络训练4 H1 A8 k2 ^5 l, {, \, I
    
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]
    8 a\" {7 H1 @. T8 h3 K4 |4 c6 v
22. net.trainParam.show=1000;
    + S/ g8 U1 l5 H0 ?( f; P% J4 q7 |
23. net.trainParam.Lr=0.05;3 c) v/ G. R6 G# c3 X1 W
    
24. net.trainParam.epochs=50000;\" \/ J' I6 |6 Q\" l- W
    
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]
    ! S/ e  P  `\" d: @- `
26. net=train(net,pn,tn);2 e& P) Z, m0 r- Q! p
    
27. 
    ! ^# w# s& N7 n6 U1 B
28. %利用原数据对BP网络仿真
    , T\" w/ [: L( x; C  W
29. an=sim(net,pn);
    ) ?: ^4 l/ j3 a$ e+ w7 A; C( B
30. a=mapminmax('reverse',an,ps2);$ d  J7 E. p: v, f2 U
    
31. 
    ( @/ w9 I* F\" K) z4 @1 V
32. %仿真结果与原数据对比测试 2 G( S4 s$ Y4 l
    
33. x=1990:2009;
    : @6 i7 k# f: J5 k  l( T
34. newk=a(1,:);\" G( L% G\" M; B9 K7 P
    
35. newh=a(2,:);3 E- |3 j! Q1 p7 S# u2 O1 E, y
    
36. figure(2);( e  H& p) x$ c! y( c7 N. m
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');1 \  q- \+ P0 |2 U  ?2 ^
    
38. legend('网络输出客运量','实际客运量');
    % f+ p0 d* T' L! p1 o$ f4 R
39. xlabel('年份');ylabel('货运量、万人');
      c  n8 {\" ]9 \% ]+ Z
40. title('运用工具箱客运量学习和测试对比图');. m/ }1 @% R# Y2 e' v
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');+ x1 @4 m7 W  X, g  [
    
42. legend('网络输出货运量','实际货运量');
    3 q9 G7 }/ F3 T- _+ m
43. xlabel('年份');ylabel('货运量、万吨');) Z$ ^/ V\" M9 [. I$ O- Z( J
    
44. title('运用工具箱货运量学习和测试对比图');
    5 M! d' e) \7 `: f$ Z+ q
45. 
    0 A. N5 j# h4 S* d# u( X
46. %新数据仿真
    + `3 g* x# k2 v( y2 @0 f6 [
47. pnew=[73.39,75.55- z' k. O7 e! `3 s- ]- `/ U) |
    
48.      3.9635,4.0975
    ) Q6 U# A- U9 x; O3 G. s! K
49.      0.9880,1.0268];( _( h% e: w0 {; Q\" P+ n$ C
    
50. pnewn=mapminmax('apply',pnew,ps1);% v\" g. ^\" s. _0 x, `- g: l
    
51. anewn=sim(net,pnewn);
    3 I. ^) v' K+ P2 D/ _8 f/ R
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. clc8 H0 n. i\" N! q; ^7 n$ z
   
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,.../ V( V$ a$ w/ |# t( N
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];5 h: r\" \& T1 q# U\" ~& {. I, y
   
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,...9 F6 d# Y) x, N4 }/ n8 R
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];3 J8 f; x, G6 U1 I. W+ B! a  @
   
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,...6 a9 R: I4 }1 D: J/ P( Y4 p
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   9 i: x5 c) I( H; U7 g6 C
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   8 z. L/ w& V1 L& j+ o
9.      22598,25107,33442,36836,40548,42927,43462];  z$ m/ D\" u  ?/ E
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    . X& n$ o; Y2 q4 `5 I5 E
11.      13320,16762,18673,20724,20803,21804];* A; J; _$ y% n- [
    
12. p=[sqrs;sqjdcs;sqglmj];
    9 T+ d5 y& Y9 j- @9 ]\" k
13. t=[glkyl;glhyl];1 T9 e0 p5 q9 `. s1 N4 q
    
14. %数据归一化
    % J/ w# g( H4 z$ w0 x9 h% l/ s
15. [pn,ps1]=mapminmax(p);
    2 i- H8 M: L- g  {! u' {+ l
16. [tn,ps2]=mapminmax(t);- ]\" ?: A. H# l
    
17. %dx=[-1,1;-1,1;-1,1];2 O4 U7 s$ r# G1 H
    
18. p0=minmax(pn);t0=minmax(tn);* g) @9 L+ h% U! J2 U) m$ q
    
19. 
    \" J9 B! G! O# w, J) p
20. %BP网络训练& `0 p* Q) ^8 C' T
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
    ! Q+ D2 [. Q# t$ g  W9 g1 A
22. net.trainParam.show=1000;
    ' G* o- j- t' h4 }2 B! w/ Q
23. net.trainParam.Lr=0.05;( H+ Y1 A/ G+ r- V& e
    
24. net.trainParam.epochs=50000;) K* Y0 J4 [( p2 d$ N
    
25. net.trainParam.goal=0.65*10^(-5);. H9 k/ X- C/ Z5 s6 V\" D- P: O
    
26. net=train(net,pn,tn);
    + C8 O& W8 z# [* ]+ w- c
27. ) a; f7 S0 r% _2 \1 z
    
28. %利用原数据对BP网络仿真
    2 P5 l+ @\" F/ K
29. an=sim(net,pn);' x* y. ^/ b7 x$ P( x1 Z. m
    
30. a=mapminmax('reverse',an,ps2);
    6 Z- w8 L0 V' J* [3 i\" W9 t
31. 
    + p( v, i/ R& M; F. a
32. %仿真结果与原数据对比测试
    * M' }5 E  I4 m0 s
33. x=1990:2009;' Q) ?( D3 n+ V% _% t9 R( L# A
    
34. newk=a(1,:);! T3 Q0 y7 W3 N  ~# C5 T\" [, L6 i
    
35. newh=a(2,:);0 l! u: \. ]$ S
    
36. figure(2);
    : t' ]% D  y8 {7 _8 `
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');) J+ s4 k, \/ L) O$ H+ r' \- u
    
38. legend('网络输出客运量','实际客运量');
    0 l9 L9 J, }( ^. A4 C
39. xlabel('年份');ylabel('货运量、万人');
    8 l+ W7 W; b% T, n
40. title('运用工具箱客运量学习和测试对比图');
    ; i/ |6 U% d# u4 ]3 U: Y
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    , p- Y7 Q' s- [0 O* a
42. legend('网络输出货运量','实际货运量');
    1 v+ N! H& l) P5 E9 ~
43. xlabel('年份');ylabel('货运量、万吨');# e$ Q% j8 [- ^: C9 B4 r1 B5 }
    
44. title('运用工具箱货运量学习和测试对比图');
    6 V% q0 U\" \. q- f+ ^2 D8 J
45. 
    - J/ f+ P2 r9 X2 h. y
46. %新数据仿真
    % O4 e5 ]0 @: x- R8 |7 |  ~! z
47. pnew=[73.39,75.55: r8 @( A. O; [' s3 x8 q* S
    
48.      3.9635,4.0975
    , {1 x2 W% E2 x$ v
49.      0.9880,1.0268];8 p\" D  Q3 |3 t: b: z
    
50. pnewn=mapminmax('apply',pnew,ps1);) ]' }& q5 K' @  [( i  h( u: S
    
51. anewn=sim(net,pnewn);- |; ~- c, I' x\" ~* @+ J0 ~
    
52. anew=mapminmax('reverse',anewn,ps2)

复制代码

T-Eric

且生 发表于 2014-7-21 18:22
- f5 I; K4 B7 }2 ?4 A/ j0 ~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4 z; ?* G, f, \0 s: r# M
   
2. net=newff(p0,t0,6,{'tansig'},'traingd');
   - q% d: g/ J: d% P
3. net.trainParam.show=1000;
   1 y) p' e. L6 ]4 w\" `/ U6 w' r8 }
4. net.trainParam.Lr=0.05;4 K% D8 t1 f% M  |; [, R5 E
   
5. net.trainParam.epochs=2000;  s4 _, t: b7 j\" _1 [# m9 d% z
   
6. net.trainParam.goal=0.65*10^(-4);
   * t! g  K' T: k% [- Q! s) z* ~1 ?
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
# E2 n- }5 c! {* z/ r* Z二、训练函数改成梯度下降BP算法 traingd
三、迭代次数改成2000
. V# W( d# U" F7 O: L- ^" c0 P上面的参数是自己试的，我也不知道为什么。
5 N# k0 N+ U3 z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