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！！
- [) z4 U# P4 V' h+ c9 i8 B
%原始数据输入
1 w# O- [0 h4 z8 `: _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,...
6 z: e, K  |  L% \8 \+ v* m4 `5 V, V    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,...
, @6 ]. M2 U: y$ X' D' S  _    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,...
5 K- p# L2 l  \: P( K4 V    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];
7 f2 W6 c7 e& r6 |8 L2 Up=[sqrs;sqjdcs;sqglmj];
+ |4 ~: e6 M8 \t=[glkyl;glhyl];

%数据归一化
  n5 c" S0 W% Y4 v8 x; Q8 }# w3 H[pn,ps1]=mapminmax(p);
2 H7 O* E2 B4 q- a[tn,ps2]=mapminmax(t);
dx=[-1,1;-1,1;-1,1];

%BP网络训练
7 F$ G' J6 ]# ?net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
( s+ I6 I, m  u1 j. c6 W7 {7 Wnet.trainParam.show=1000;
net.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
: K$ k* h8 o8 w5 f: Unet=train(net,pn,tn);

%利用原数据对BP网络仿真
an=sim(net,pn);
$ D' d$ x1 H4 u5 }. Ca=mapminmax('reverse',an,ps2);
8 I8 @6 L7 _5 ^
%仿真结果与原数据对比测试
+ P( X- Y& |2 yx=1990:2009;
newk=a(1,:);
newh=a(2,:);
figure(2);
% ~! v1 K4 g  Usubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
- {  v* z& r4 |5 t$ ^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');
) Y/ z2 d4 y; |; M) J* vsubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
% {+ ~! }1 d: b1 q. G- Llegend('网络输出货运量','实际货运量');
2 H$ x7 U2 E1 o5 g, J# w: Rxlabel('年份');ylabel('货运量、万吨');
# N9 X! L) U: C, s$ |5 Gtitle('运用工具箱货运量学习和测试对比图');
2 y$ ~$ S, w0 ?' a- `
%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);
$ Y* h; P* ]4 ]5 Ganew=mapminmax('reverse',anewn,ps2)
3 i1 c! s3 {* h0 n6 _6 N- r2 ~
' B9 L# k5 H. T# H2 [

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 ...

# i( o! H4 Q' x! ~" R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：
8 a- O1 d% m6 e( Fnet=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：
! c9 a, }! T3 k$ z/ x+ `+ X6 `2 D[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
a=postmnmx(an,mint,maxt);
; D& r: q6 J; S6 T- I* ^) epnewn=tramnmx(pnew,minp,maxp);
, J, `" S- ^; ~5 w* z& T- \1 Kanew=postmnmx(anewn,mint,maxt)
我修改为：
[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
: M1 A. K: H% x; h  Ya=mapminmax('reverse',an,ps2);
- E0 [5 m( f; d. ]$ @pnewn=mapminmax('apply',pnew,ps1);
anew=mapminmax('reverse',anewn,ps2)

% k; w; p% u# J  _原程序为：
%原始数据输入
clc
; |1 z% s6 a6 a3 S1 Z! b1 Vsqrs=[20.55,22.44,25.73,27.13,29.45,30.1,30.96,34.06,36.42,38.09,39.13,39.99,...
9 V3 x6 h! K7 ]8 A    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,...
- e  K/ U% ]) ~( ~' L    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];
) B9 z2 v0 U5 k& d- ~& H/ _glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
3 T$ i2 d- i' e    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
9 P1 h/ q$ }: }0 T) Kt=[glkyl;glhyl];

%数据归一化
[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
dx=[-1,1;-1,1;-1,1];
: p+ y' j: e; m! I8 d4 Z9 I
%BP网络训练
net=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
7 a* r# j% K& K  r! n. }net.trainParam.Lr=0.05;
0 f# r2 {+ w9 A- snet.trainParam.epochs=50000;
9 N. o1 Q8 g3 i0 x" D& h5 G6 g8 Onet.trainParam.goal=0.65*10^(-3);
2 f0 H- ]$ I% I' j& Y9 ^( ~0 Fnet=train(net,pn,tn);

%利用原数据对BP网络仿真
an=sim(net,pn);
7 k0 ?$ ^. x7 a% S! Oa=postmnmx(an,mint,maxt);

: K' I/ G" j% G! e%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,;
2 _/ A) W1 i. g. E: {newh=a(2,;
5 e" T3 K5 w8 |( C# U3 ~  Xfigure(2);
( `# z' z9 O! t' v: Ssubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
legend('网络输出客运量','实际客运量');
, t: l' X/ B- c+ x( Lxlabel('年份');ylabel('货运量、万人');
% n* b. _  v2 |) t' |5 k8 s8 Etitle('运用工具箱客运量学习和测试对比图');
: P7 b9 ~2 Z7 |3 ~! p% ?$ ]: O6 d& asubplot(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');
- w! t) F$ r  K5 a$ S8 A* P& b) Ftitle('运用工具箱货运量学习和测试对比图');
2 X/ y2 |$ {8 M' ?( S7 L* O
, K! L, [) R* G9 a, z7 {2 c%新数据仿真
9 _& [! O( `1 wpnew=[73.39,75.55
    3.9635,4.0975
    0.9880,1.0268];
# T7 A8 ^# t. H) H! \6 apnewn=tramnmx(pnew,minp,maxp);
; \. x. N" r+ x9 F4 m& H* Banewn=sim(net,pnewn);
2 b/ [0 Y/ S* ?8 c" Z# \; Fanew=postmnmx(anewn,mint,maxt)

9 N1 A7 C8 T! C8 ^' q( M1 p% l修改后程序为：
%原始数据输入
% U, B9 G6 a8 f8 q' Q3 h* lclc
( B0 S' i. A0 {4 N) r: |( Asqrs=[20.55,22.44,25.73,27.13,29.45,30.1,30.96,34.06,36.42,38.09,39.13,39.99,...
1 ~6 d7 c, A7 x" i9 J+ K    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
  Q; f3 I+ c# Zsqjdcs=[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,...
& t. y, C0 C5 ^( A) a" b8 q    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,...
7 v, \& z& Z, E: S& x& b" e    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,...
  U- T  \. c6 m6 y    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];

) s+ q/ \* D8 R" j& C%数据归一化
1 e! G" k+ N  i1 W7 X$ N1 f8 Z[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);

' I. f) c) l- @- r%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
, Y, ]* c" m! x' Y9 @% b+ l% Wnet.trainParam.show=1000;
; S4 q' l: q7 l+ ^* H! P5 Znet.trainParam.Lr=0.05;
/ y5 ^. J# m1 Snet.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
5 R1 ~! H. Q0 o4 R0 o: [. Ynet=train(net,pn,tn);
$ u, V2 @$ h. h% D- z& a; N$ H, A6 e
' x, \4 l" y* k- Y%利用原数据对BP网络仿真
8 t, j) Y- B" y" Q! O! O) Y+ Gan=sim(net,pn);
a=mapminmax('reverse',an,ps2);
% j8 F6 X( v8 j$ I+ y$ `- l, p2 T/ s
%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,;
newh=a(2,;
- T* _; ?* X4 K% ]6 M/ Y& ofigure(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');
* \- R+ q+ T. a" A# r' \6 G7 h6 osubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
) ^2 m! X4 k9 G, Wlegend('网络输出货运量','实际货运量');
xlabel('年份');ylabel('货运量、万吨');
. k9 A  O# g1 J4 B2 J7 J# |title('运用工具箱货运量学习和测试对比图');

, F: Y7 f- z/ F6 p( _# \" P, N4 }%新数据仿真
+ v3 e& k, t  f5 u" [1 Cpnew=[73.39,75.55
1 {* i" _2 D! U( k    3.9635,4.0975
    0.9880,1.0268];
- e* m, }* Z' P8 P+ O: Qpnewn=mapminmax('apply',pnew,ps1);
4 c% ~$ u# v$ [$ @  r1 ?( m% \anewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
2 ~/ z; I" a4 E; e/ X& m/ K（修改的地方用颜色标记了）
. z# ]0 [- t4 S; G: \1 r9 s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
   ; U3 A: C) E- V# x4 j) c! {- X( G4 v
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 ~. y) d, R9 f, z( ]
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];- U: P8 S8 d+ D; s. ?
   
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,...
   4 K& l; ~! q' g# Y$ ^! U, R
5.     2.5,2.6,2.7,2.85,2.95,3.10];4 R5 k9 P8 s3 x: V
   
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,...
   $ v! }, n) s7 n4 S! R* P! u
7.     0.56,0.59,0.59,0.67,0.69,0.79];3 `8 j  n& n3 j0 _$ l. h
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...' r! P6 ?* }3 k. b/ \3 j
   
9.      22598,25107,33442,36836,40548,42927,43462];+ [& {% \7 F2 S
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...5 `$ q( D6 R) P* }  L2 Y+ ]
    
11.      13320,16762,18673,20724,20803,21804];& o, B0 m# i5 z! n* k$ d
    
12. p=[sqrs;sqjdcs;sqglmj];. N2 A# I8 v& r\" N, g
    
13. t=[glkyl;glhyl];
    ) W! g; ?' V; Y/ N* E8 W- o) r) P
14. %数据归一化0 j! |5 a- W! l- a, P
    
15. [pn,ps1]=mapminmax(p);3 G5 O/ |$ J0 ~  y3 Z
    
16. [tn,ps2]=mapminmax(t);
    ( o$ ^$ ~2 [$ Z% I( v+ M
17. %dx=[-1,1;-1,1;-1,1];
    . Z2 v/ Y- m/ Z) l- v; O4 W
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]- x; u9 U6 ]* t  a: T5 G( X
    
19. 5 q- C4 L0 I4 i
    
20. %BP网络训练* K( d6 L\" D* O6 l6 {3 ]3 e
    
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]2 C+ Z2 F- P, G  n0 u\" W3 f
    
22. net.trainParam.show=1000;% m, E. z6 x) z. r! s& G
    
23. net.trainParam.Lr=0.05;
    - Y, R  H4 U. O9 y6 W
24. net.trainParam.epochs=50000;* ?+ j! K7 D/ w. ^
    
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]2 b$ h6 {+ O\" v( ]\" b1 c
    
26. net=train(net,pn,tn);
    : P\" T4 x/ W8 d: f$ `5 I9 _
27. 
    ) o' b! ^0 B# k/ I9 Y! ^2 f+ f
28. %利用原数据对BP网络仿真
    7 R5 m2 r) E* ]' b
29. an=sim(net,pn);8 M0 Y5 ]$ c. l( V# X% k
    
30. a=mapminmax('reverse',an,ps2);
      [6 d' k' Z# C  @5 Q3 l
31. 
    1 y+ N% M+ u0 r$ q
32. %仿真结果与原数据对比测试 3 X6 q8 y! Z! O4 d7 u. j: f  l
    
33. x=1990:2009;
    ) C  a$ R+ |: k: Z
34. newk=a(1,:);
    & U; ]- {% c5 q
35. newh=a(2,:);
    ; g! N) l! ^0 L/ H* R3 a
36. figure(2);6 d5 c\" Q7 {. M* y8 b9 Q( j
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');: ?\" V% s1 G  q5 _- F
    
38. legend('网络输出客运量','实际客运量');! x( m+ B% C( g  z\" u
    
39. xlabel('年份');ylabel('货运量、万人');
    6 N1 G; B& b/ b# `0 D
40. title('运用工具箱客运量学习和测试对比图');$ ]8 u1 x7 a6 W2 Z) P
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    $ p! X6 e  b9 f! f$ f* \+ u
42. legend('网络输出货运量','实际货运量');
    & t) i# J( g1 c# r! A* }
43. xlabel('年份');ylabel('货运量、万吨');3 O. q, c* Y( e# P& t' |0 f
    
44. title('运用工具箱货运量学习和测试对比图');5 t0 L/ I# V# K: m% N
    
45. ( M% a7 N\" G/ N( W- ]/ N2 c# S
    
46. %新数据仿真& I- h- G- N# I\" R) S
    
47. pnew=[73.39,75.559 G* o) ~\" U5 X3 m! t6 L4 g
    
48.      3.9635,4.0975$ r1 E  B7 {/ T& b3 b: _
    
49.      0.9880,1.0268];* |6 Z6 d. e9 Y& s8 x) |8 @* L( y
    
50. pnewn=mapminmax('apply',pnew,ps1);8 v0 W! M6 q9 H0 m! c1 |* c4 I\" B+ C
    
51. anewn=sim(net,pnewn);
    6 |! c/ I9 S4 a3 @0 M. x4 z$ d0 G
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
   / I& P0 A8 s: U, W* }\" p; 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,...
   & B\" _6 ]1 n1 V, e& \- m
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];/ K/ [$ `% F3 i. D: r9 |
   
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,...- w$ }' K5 V9 d, D+ d. L1 o
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   9 u' V8 N\" ?( ?5 L, h  u
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,...; n! }: V! o. n0 ?
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   ( t/ K# G7 N\" y\" u- Z: S
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   / p  h3 M8 z% t& X
9.      22598,25107,33442,36836,40548,42927,43462];
   ' @: {( ]& n+ r  ^+ o1 Q2 V& z7 C
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    ( w\" p+ l- P+ j0 ~- L; Q9 U
11.      13320,16762,18673,20724,20803,21804];
    2 o5 ^: t5 A# E/ _/ p' @
12. p=[sqrs;sqjdcs;sqglmj];' ^* Z8 w4 J' q  |
    
13. t=[glkyl;glhyl];
    / K. ~! E$ ?7 m% _
14. %数据归一化: }+ ~1 H+ Q1 H0 Y\" j6 u' u3 q
    
15. [pn,ps1]=mapminmax(p);) X, s- L6 `1 l8 F* t& @1 f: m
    
16. [tn,ps2]=mapminmax(t);$ J: _( D, Z' ~4 ^+ f% m9 H$ Q) m$ _
    
17. %dx=[-1,1;-1,1;-1,1];% c\" p, w  T. G$ P. p
    
18. p0=minmax(pn);t0=minmax(tn);' w7 c6 ~, \6 m% j! g
    
19. 
    5 \1 f: {\" S' M% E- k/ x1 z! A
20. %BP网络训练+ F* z3 `% w( e( P  R/ N
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); # _\" j  Q! P$ V/ g
    
22. net.trainParam.show=1000;& k3 I' g. }' u5 X! l: L
    
23. net.trainParam.Lr=0.05;# o3 P0 J* _1 x3 ]7 u
    
24. net.trainParam.epochs=50000;
    4 T; I# Y* e; B# @- C) e
25. net.trainParam.goal=0.65*10^(-5);
    \" l. Y/ V. P4 o; `\" K8 S5 N) h
26. net=train(net,pn,tn);8 H* ~( R% K5 @+ v* _# j
    
27. 7 b* }- y% g9 z) w) `\" C6 k0 ?
    
28. %利用原数据对BP网络仿真
    # G) V, V: B3 w' \  T\" d0 Q
29. an=sim(net,pn);
    # l# `3 Q7 N' x/ g
30. a=mapminmax('reverse',an,ps2);$ ~% V: f) D& D
    
31. 
    6 I5 M/ Y+ f! E2 Y/ p. P
32. %仿真结果与原数据对比测试 7 y; D! e\" q, P+ U
    
33. x=1990:2009;
    # |8 n7 V6 Z' J* G8 a
34. newk=a(1,:);( `: t1 \- w7 P6 b
    
35. newh=a(2,:);* S\" s! R7 g* _6 E3 v) n( q
    
36. figure(2);4 l# _' f2 h+ }/ y  J- g) S
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');) E* p) G* F: o6 O7 S1 p
    
38. legend('网络输出客运量','实际客运量');6 ~/ ?. j2 f- v. d0 g% E
    
39. xlabel('年份');ylabel('货运量、万人');/ L* H# w6 J: O% U
    
40. title('运用工具箱客运量学习和测试对比图');
    / E5 ^& |2 R& Q/ Z% d; m/ g' ]
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');' `\" r9 I# u+ g' D) V. U
    
42. legend('网络输出货运量','实际货运量');  \9 L! Y1 v/ w+ _6 n6 g8 I
    
43. xlabel('年份');ylabel('货运量、万吨');4 K% ?: Z4 |% ^1 N. k/ y
    
44. title('运用工具箱货运量学习和测试对比图');* m3 F! m' n8 m- B4 r. f: w
    
45. ( y, T. I2 v; f) _# z
    
46. %新数据仿真
    2 D( s4 s  T3 y3 U
47. pnew=[73.39,75.55% o\" l0 F\" m. o\" ^. p$ \, p
    
48.      3.9635,4.0975
    2 f, ?4 x5 C7 W: ~
49.      0.9880,1.0268];, ~: {& L* ?: [* a
    
50. pnewn=mapminmax('apply',pnew,ps1);( W' U% g6 x5 E& y( u
    
51. anewn=sim(net,pnewn);
    % D& y  d+ Z0 l
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

* L" G, v9 K& j* A5 U& v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 n! M\" D( v  W/ I' @
2. net=newff(p0,t0,6,{'tansig'},'traingd'); & R2 n# T' y& F+ i% z& V$ {4 D: y
   
3. net.trainParam.show=1000;/ F) Y; d\" e/ Q\" A$ F; H  L2 }8 l
   
4. net.trainParam.Lr=0.05;. {2 v8 E& x$ L- _1 j
   
5. net.trainParam.epochs=2000;
   8 K9 D# h6 q' E* \+ i9 e$ i
6. net.trainParam.goal=0.65*10^(-4);* _0 s\" w+ F# m$ |
   
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
' w, K0 T$ z$ S0 c) r$ S二、训练函数改成梯度下降BP算法 traingd
) ?  _$ h/ @$ @! _7 S6 M/ \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