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！！
! b. h2 d" g- w4 h1 h
2 E' J2 I/ o$ d+ X%原始数据输入
  v, ~7 }& i0 i' C$ K  d) ]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,...
  i( C; s5 u3 d8 u" e8 @& j    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,...
: y- O* I  W7 m$ Q: S: f    2.5,2.6,2.7,2.85,2.95,3.10];
0 s+ Z4 C& U( T/ |8 s  C/ usqglmj=[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];
# e+ `" {$ ?0 |3 M  S8 oglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
4 I- q, l0 M7 r+ T! m    22598,25107,33442,36836,40548,42927,43462];
* Z; U& |+ H/ I5 G% uglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    13320,16762,18673,20724,20803,21804];
! ]1 Y2 F8 h2 ap=[sqrs;sqjdcs;sqglmj];
, v3 R5 u: h7 t5 m- ]t=[glkyl;glhyl];
8 M1 n1 a3 i, X1 E
%数据归一化
[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
dx=[-1,1;-1,1;-1,1];
0 V% ~2 p1 ~9 ~2 {$ L3 P4 N0 l
%BP网络训练
! i* S9 j. v6 T5 r: A5 B: tnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
& ^4 D7 I& h( R5 o: D4 J# I/ Rnet.trainParam.show=1000;
  A% Q6 ^2 J+ {: knet.trainParam.Lr=0.05;
0 i+ J9 e3 @$ D) U$ D5 Unet.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
; {& _. y* v- _/ Q( Znet=train(net,pn,tn);
1 z" x* m3 p$ g5 s& L- e# T6 ~$ c
%利用原数据对BP网络仿真
0 S$ F3 r) a4 B: ^; J$ @an=sim(net,pn);
) Z: Q. B2 ?' J: t; r/ Ca=mapminmax('reverse',an,ps2);
' f" n3 O% b( a# H3 X
  E! _3 ?3 H$ `7 L%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,:);
5 v" M0 Z4 u" k# T8 s2 `* Knewh=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');
0 K7 u& J$ u: M; ztitle('运用工具箱客运量学习和测试对比图');
# q0 O8 t, c' r! P4 Jsubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
7 i$ M% }4 c" a& r% Elegend('网络输出货运量','实际货运量');
xlabel('年份');ylabel('货运量、万吨');
, ~4 M3 A! M& S8 B/ s- m2 Dtitle('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
    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)

8 w) W" [' g- T$ m( S, T+ s

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
: w7 z: z& i+ j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 T/ [0 B' p- x: |% N$ Q) nnet=newff(dx,[3,7,2]{'tansig','tansig','purelin'},'traingdx');
我修改为：
1 Q9 b+ J& j, B7 _net=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);
- X, S6 u" x  p. Z) g2 F0 q4 Y8 {a=postmnmx(an,mint,maxt);
' R: l7 \5 z! v% h7 u3 S: V. Ipnewn=tramnmx(pnew,minp,maxp);
( l8 Z# ?2 G1 i5 J" fanew=postmnmx(anewn,mint,maxt)
我修改为：
3 z7 Z) R: F2 h) @# R6 Z5 ^& m/ |[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
1 a: F; G1 ]  Ea=mapminmax('reverse',an,ps2);
pnewn=mapminmax('apply',pnew,ps1);
anew=mapminmax('reverse',anewn,ps2)
$ F: |  w# ?/ ^( n4 F$ l+ @
& R: \1 }4 v& y- r) L% g原程序为：
%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];
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,...
' n4 b+ B6 ?( o6 G  p+ ^1 ~    2.5,2.6,2.7,2.85,2.95,3.10];
1 `  ~( y# h: T# ~8 H2 Qsqglmj=[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];
+ Z9 I- m& h/ ^) \glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
- Z. E2 R, B! m4 R3 x    22598,25107,33442,36836,40548,42927,43462];
0 E& O5 _0 M& k5 u3 `; Zglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
- c* J3 T3 e' n: b+ H* X; k6 ]: It=[glkyl;glhyl];
+ G- h2 j- W3 Q, {  n+ x' ~
%数据归一化
6 B$ n- a  G# w5 Z[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
# w! K& X* F! A; Wdx=[-1,1;-1,1;-1,1];
" H1 J' o6 v' z8 k% \: H  d
%BP网络训练
' B/ q8 h. ?* c2 K: G' b# knet=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
& D# J, `0 S( ]4 y+ b9 I5 |net.trainParam.show=1000;
net.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
% D9 b- D8 R* T/ }' H- J3 Jnet.trainParam.goal=0.65*10^(-3);
7 n& I, Q2 Q. ?' mnet=train(net,pn,tn);
  \# U0 b$ E$ ^' W" r3 t. ]- ]
  r& ?( s' k5 ^% n. Y7 H2 V%利用原数据对BP网络仿真
an=sim(net,pn);
; b, q  J& N/ D6 \) |# ?a=postmnmx(an,mint,maxt);

%仿真结果与原数据对比测试
x=1990:2009;
/ v% p+ i( b& m' V3 |* Dnewk=a(1,;
newh=a(2,;
figure(2);
3 L8 g8 ]% l% ^0 S4 \" B- Psubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
6 A) B; H& n9 m0 {6 v& C$ }$ x0 Ulegend('网络输出客运量','实际客运量');
1 T: a, U8 i$ H, z' 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');
) G7 k5 A0 j9 g4 M  Dxlabel('年份');ylabel('货运量、万吨');
$ m/ `9 N7 N3 V# G2 O8 U6 ?title('运用工具箱货运量学习和测试对比图');
4 n# [6 U. ]$ t! S- R: S/ @% Y) ?
%新数据仿真
4 p5 i9 R! k$ g7 K& Xpnew=[73.39,75.55
# _2 r" m( Q. B    3.9635,4.0975
* p) {1 }- Y; T) Q    0.9880,1.0268];
( h& M! [- i5 n: C. V% s; Bpnewn=tramnmx(pnew,minp,maxp);
anewn=sim(net,pnewn);
- ~4 k, V4 x: l- F3 @' B) [: i& ?# I- `anew=postmnmx(anewn,mint,maxt)
. b  L9 C" B2 T1 O+ }9 ]
修改后程序为：
2 I  f% R' N$ R  l! ?%原始数据输入
clc
- t+ q7 k- ?* f! S7 M# Gsqrs=[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 t' C* E. F8 h1 i/ B    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
' F5 V& W7 I. y0 u. U4 T" Dsqjdcs=[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,...
+ n; c# y4 {- I* m    0.56,0.59,0.59,0.67,0.69,0.79];
: [9 i# u3 F6 gglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
4 i; M: ]3 Y  {8 s( Z8 y" V) e3 b    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];
3 K: e4 r$ |9 Mp=[sqrs;sqjdcs;sqglmj];
3 s# _1 e2 e: g* I- `9 Vt=[glkyl;glhyl];

%数据归一化
: d6 r; r4 G+ P+ t, C. {9 ~* s[pn,ps1]=mapminmax(p);
! `3 x; L; A' h4 \6 Z# `[tn,ps2]=mapminmax(t);
: q* {2 F1 u( x3 V
% G0 ~& b1 c3 N; y8 ?- V5 @%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;
2 g1 ?) I" V9 ]( i+ ]net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);

%利用原数据对BP网络仿真
an=sim(net,pn);
a=mapminmax('reverse',an,ps2);

+ [/ ], g  Z2 X% z% F- k2 G, [%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,;
newh=a(2,;
9 N! n; ^" j. U$ E' Q0 cfigure(2);
7 M) J  r, M8 K6 |3 P& m1 K: d. |subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
* q. z5 Y& {1 R+ ]$ a* e+ U1 D9 vlegend('网络输出客运量','实际客运量');
xlabel('年份');ylabel('货运量、万人');
; r0 E6 W1 T! ^' m# ftitle('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');
7 M/ Q/ C& t0 G" }9 lxlabel('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');
8 M# N- ~7 ?' Y1 r
%新数据仿真
; H) f( j! ]+ }% Hpnew=[73.39,75.55
/ j' \: u2 n" }9 Q4 y7 M    3.9635,4.0975
    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
* w1 q& a8 S/ Wanewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
' X% Q$ L. Z: w% q! ^（修改的地方用颜色标记了）
. l& j/ @# _" p* B( K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. clc9 {/ \5 G! b( U\" J  |
   
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,...
   # a, w. t' ^- x# h% [
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   7 s( ?5 g) B1 h0 G1 p& }+ 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,...( [9 j/ h. q( h- z1 b% P5 C. B
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];7 j) ?) o  A1 W( S$ h# t+ \$ X
   
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,.../ K0 N3 Z\" Q% D/ Q5 x, _& v) n6 D: E
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   7 t7 h1 }! h7 |
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   % A, K7 ?8 H4 A* u1 }, o
9.      22598,25107,33442,36836,40548,42927,43462];' C\" F3 X: ]8 l4 `0 m; A1 i' F
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    4 ?2 E\" H  C2 p\" \7 _6 j
11.      13320,16762,18673,20724,20803,21804];9 u- l0 V4 F( J! U/ |. k
    
12. p=[sqrs;sqjdcs;sqglmj];; t& E6 S% p) G' z5 ^, Y; ?
    
13. t=[glkyl;glhyl];
    1 ~( }6 W: \, G; s9 \' O2 F! N
14. %数据归一化
    7 _) t3 t5 y+ m0 B\" E
15. [pn,ps1]=mapminmax(p);
    # P1 T9 ~: y5 L4 V8 C4 e' I
16. [tn,ps2]=mapminmax(t);
    . `8 L\" C; o1 L! N+ G9 {
17. %dx=[-1,1;-1,1;-1,1];
    $ g4 j6 a* P. B  k: M( a
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]
    & j/ X2 u8 f- H0 T
19. 
    & ?5 x: c: A0 _
20. %BP网络训练
    : P* |0 z% W7 Z9 s3 j8 Q9 W
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]6 F# f2 D5 i6 n( m5 `7 U( w
    
22. net.trainParam.show=1000;$ T! C\" \0 X9 i% p. `4 d
    
23. net.trainParam.Lr=0.05;% d# }, n8 l- x& V7 Y8 B
    
24. net.trainParam.epochs=50000;
    8 S+ w4 K7 u7 `/ m, u  O
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]* [+ q  a5 g/ j- k4 Q0 C  t0 S
    
26. net=train(net,pn,tn);6 `! b$ R1 U: n1 D
    
27. $ y0 S( U9 E- E& `- M1 \' S4 I
    
28. %利用原数据对BP网络仿真: T( K/ _' L9 I, [6 n& p- Q) W
    
29. an=sim(net,pn);
      e( \0 ]9 Z) j/ v
30. a=mapminmax('reverse',an,ps2);1 |7 J. I. h$ Y
    
31. % e. {* l& U5 e\" \. |+ m7 W7 Y4 j$ p
    
32. %仿真结果与原数据对比测试 % _- A4 {& Y8 O, y% u
    
33. x=1990:2009;
      O- `5 V; |' v  X+ m) C% X
34. newk=a(1,:);3 u\" ?0 a& E' V4 _6 \- P6 x3 D
    
35. newh=a(2,:);
    ! T2 s\" o1 w# z, B6 n
36. figure(2);0 c\" @) }: t, t! }: ~% K1 b
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');9 x2 J' o& f: h! S
    
38. legend('网络输出客运量','实际客运量');
    3 I& |  [9 h  X% U2 n
39. xlabel('年份');ylabel('货运量、万人');
    & j9 r) X) c- p4 W! Q
40. title('运用工具箱客运量学习和测试对比图');) i( d* }: r: t+ c: @
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    $ ^- e* s\" V, g
42. legend('网络输出货运量','实际货运量');
    9 _$ o\" N1 h2 Y3 z$ p/ n
43. xlabel('年份');ylabel('货运量、万吨');! d, w3 h, s9 z! Q+ R7 l7 }( H
    
44. title('运用工具箱货运量学习和测试对比图');
    : E' ?% y+ [0 C* k+ \) s4 ^
45. / E5 b) U. D/ ^1 d2 a: c8 V
    
46. %新数据仿真
    0 h0 l! |2 T+ y0 X2 b4 `
47. pnew=[73.39,75.554 ~2 r5 T9 J8 d( m& ?8 Y' l
    
48.      3.9635,4.0975  ^4 ]* X0 o, a0 C7 @  k
    
49.      0.9880,1.0268];
    / p- W\" ~( d) T0 o
50. pnewn=mapminmax('apply',pnew,ps1);\" V# i% _7 `0 m& Z
    
51. anewn=sim(net,pnewn);; W6 D3 A0 v1 ^- u; [2 J6 ?
    
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. r) {\" L2 s8 v) F$ W9 t
   
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,...
   ! W& c* V' y2 S+ a+ U
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];3 L' _, r  A( l7 z+ O9 n8 ]; A/ r3 g
   
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,...
   * K8 @. h5 n( ^' ]+ J
5.     2.5,2.6,2.7,2.85,2.95,3.10];1 R8 f$ c4 h+ ]
   
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,...
   , q) h% T# y& j' ^  a7 o+ E
7.     0.56,0.59,0.59,0.67,0.69,0.79];; i2 l( W& d) X  W4 T
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...6 q' f% Y. x; i0 C; x* q9 `
   
9.      22598,25107,33442,36836,40548,42927,43462];0 W: ^+ O7 n; p; y1 a! ^
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...3 s$ x2 ?9 i- e; Y( q9 t\" f
    
11.      13320,16762,18673,20724,20803,21804];9 q1 O+ r3 }\" }  D
    
12. p=[sqrs;sqjdcs;sqglmj];
    6 W) K: x0 H1 v4 [- N
13. t=[glkyl;glhyl];
    ; f$ N) H  y3 M5 r. l+ k
14. %数据归一化
    0 M  x8 V$ c) j2 l$ G
15. [pn,ps1]=mapminmax(p);6 ], d# q, X# z- y$ w/ y
    
16. [tn,ps2]=mapminmax(t);
    ' P8 z9 g1 }4 O
17. %dx=[-1,1;-1,1;-1,1];
    - B5 X- \* Z  Y1 I) l3 v% @
18. p0=minmax(pn);t0=minmax(tn);( s8 a. G9 J5 V! l: W6 `8 n
    
19. 
    4 r9 Z7 Y9 z0 Z! t4 P\" p
20. %BP网络训练$ }4 B* c7 u# f4 O8 W\" U
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
    \" R- g2 j\" Q) B* n% R8 j
22. net.trainParam.show=1000;
    * o) s, r* j  q8 \
23. net.trainParam.Lr=0.05;
    & L8 {0 f/ K' y# w) i3 f; _# S5 m
24. net.trainParam.epochs=50000;
    7 D! t, x7 ]8 x
25. net.trainParam.goal=0.65*10^(-5);/ v! a( I0 C) N: ^
    
26. net=train(net,pn,tn);
    5 x* m: p- i! v1 B5 U
27. # O8 y8 T( E2 v+ x
    
28. %利用原数据对BP网络仿真
    - ~- R! s* ^* T# n: B
29. an=sim(net,pn);0 \* \! [( z  d0 T3 k
    
30. a=mapminmax('reverse',an,ps2);
    ' e2 q) T3 y: I8 Q- A' `
31. 
    ! s( G/ A/ q# G; B2 f* M) V$ }
32. %仿真结果与原数据对比测试
    3 j  @# U% v) i3 {! j. S
33. x=1990:2009;
    5 I' P$ }: e2 p. ?
34. newk=a(1,:);
    ) P1 }) o6 E  C1 V
35. newh=a(2,:);
    & Z* k5 s\" V\" C: w; k$ a
36. figure(2);; v3 Z* r% q! ~+ R9 _- }+ G  s7 {
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
      c\" u% h1 n* n0 ^
38. legend('网络输出客运量','实际客运量');
    0 P1 K  t: Z& Y$ |
39. xlabel('年份');ylabel('货运量、万人');7 j2 t' L7 R+ N+ Y
    
40. title('运用工具箱客运量学习和测试对比图');& `2 a# L0 p8 l8 M* u! T. o! |; B+ T
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');8 W7 f1 h1 O7 K6 N
    
42. legend('网络输出货运量','实际货运量');
    & D' R6 Q2 J$ T' n8 `4 v9 B* m# S
43. xlabel('年份');ylabel('货运量、万吨');
    8 n3 f* t' A1 K8 D
44. title('运用工具箱货运量学习和测试对比图');
    8 H6 N' q4 r- M4 q1 p
45. ! ]  `- r& a/ a* c+ }- M6 i: G1 k
    
46. %新数据仿真
    % f& ^& L4 I7 Q/ @! L
47. pnew=[73.39,75.55- G$ Q0 o+ S$ q' T/ c) t
    
48.      3.9635,4.09756 l5 b$ I0 K4 E0 Q
    
49.      0.9880,1.0268];
    2 g& @; b& a4 M* k9 b, i$ C$ Y/ ^
50. pnewn=mapminmax('apply',pnew,ps1);
    7 i+ F# O2 P: f% i2 h2 o
51. anewn=sim(net,pnewn);
    2 O$ E+ D: U3 X, x
52. anew=mapminmax('reverse',anewn,ps2)

复制代码

T-Eric

且生 发表于 2014-7-21 18:22
" q" e& A, p& U7 X+ q9 Z; q7 J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
   2 o; \. t: B' E/ p% N  q
2. net=newff(p0,t0,6,{'tansig'},'traingd'); ) C& `7 U3 e/ x  q. v& O
   
3. net.trainParam.show=1000;5 j  T; ^% T0 e0 D  w3 A- t1 n' e& g4 F
   
4. net.trainParam.Lr=0.05;
   2 f6 x- e! k$ e1 t: k2 a
5. net.trainParam.epochs=2000;+ t  H3 X; V/ ^. {
   
6. net.trainParam.goal=0.65*10^(-4);' f' w/ p/ W. @) Y1 a3 O6 b
   
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
/ F/ L! v4 H5 e* a- _7 w二、训练函数改成梯度下降BP算法 traingd
三、迭代次数改成2000
% d6 ?' I( ]/ t) R# i5 [# m+ C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。
6 }- D' B0 F3 _' d' A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