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！！

5 s# E. A5 F2 f( k9 O  U%原始数据输入
clc
9 f8 e6 l9 l/ T' _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];
) E# q, C$ H7 r/ M: }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,...
; m0 U! M# K5 ]7 c    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];
. g$ K. @8 ?2 r2 rglkyl=[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];
* h, y+ v% P4 w2 G9 g
8 U7 q' y# {& i0 J( P%数据归一化
[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
dx=[-1,1;-1,1;-1,1];
4 A% I* q. @/ r  l% o
%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;
9 k& Z7 @* d0 P9 j, L9 Tnet.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
3 ^. S" |$ `; A& h3 ~# enet=train(net,pn,tn);
" U  R; r+ L9 j
: x8 Q  I. V0 ~1 G9 S) `4 J3 A# S& X0 z%利用原数据对BP网络仿真
( q5 w  d2 g( c+ {( `an=sim(net,pn);
, Z3 Z% Y$ n/ l8 }) n! m6 e$ Va=mapminmax('reverse',an,ps2);

%仿真结果与原数据对比测试
( l; f+ x' ]! j+ d$ {$ X( O2 Qx=1990:2009;
$ I( ~/ L( ^/ I* tnewk=a(1,:);
& O0 Y! O3 D( S: ^5 ~newh=a(2,:);
: i. Q, j- Y* J; S6 U5 ufigure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
. B2 z' E9 m# A5 ~legend('网络输出客运量','实际客运量');
xlabel('年份');ylabel('货运量、万人');
0 o6 C0 N% H# Xtitle('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');
" ~8 v! ?7 `- K! P2 ^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');
' z' {* }0 `' I- i! L) `
. F3 k9 u8 i1 K4 t1 w8 [" m%新数据仿真
pnew=[73.39,75.55
* ~) B) _' ~) C2 L6 L  ]* T    3.9635,4.0975
2 H4 Z  j" n* h" q    0.9880,1.0268];
' ~1 x& Q. G" V% hpnewn=mapminmax('apply',pnew,ps1);
6 _) M- ^' Z6 G- M0 }2 u) D! fanewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)

+ B! b7 l; }  n+ f2 \/ a+ k! U

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
6 y- ]+ B/ c$ z% j/ C$ g: ?1 ^- Z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');
: K/ n/ [- q! x" v6 o我修改为：
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
6 w3 H6 z0 U! 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);
: c3 S2 y3 r! W# x% \anew=postmnmx(anewn,mint,maxt)
我修改为：
[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
a=mapminmax('reverse',an,ps2);
* B7 {/ z! j9 G9 \" Q9 |pnewn=mapminmax('apply',pnew,ps1);
anew=mapminmax('reverse',anewn,ps2)
+ {! |; K6 `8 H2 u+ p2 K( m
; H1 c7 B5 I- N0 L: ~- n0 k原程序为：
%原始数据输入
1 f+ J6 j9 e* F% Z7 P) o- uclc
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 {: M: y  r. j# u    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
! o( P6 v+ M, Hsqjdcs=[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,...
0 R8 q% Q) N# 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,...
6 C) ^9 f0 @; m$ u, c    0.56,0.59,0.59,0.67,0.69,0.79];
, J3 r( D) @- p2 }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];
, t# S( i8 F0 C4 bp=[sqrs;sqjdcs;sqglmj];
+ d$ f/ s# C* X% P7 _& j# |$ Yt=[glkyl;glhyl];

5 g. y  ?4 E. U6 f8 h%数据归一化
$ v+ E. M# g* O4 N4 Q6 a  b2 `[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
: s  H! M4 W8 m: i: B, Q$ Pdx=[-1,1;-1,1;-1,1];
  O# _- Z4 p$ |; {  ?  g2 }
1 \' e: _, u8 N1 X% u6 R# T. r%BP网络训练
* ^4 F' Q% c: n8 K# {" ?# C  w7 Q: ynet=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
net.trainParam.Lr=0.05;
1 m  F2 k5 {" h# Gnet.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
: j6 d2 x7 ?! x' d6 Q/ X' znet=train(net,pn,tn);
5 @% ]- u9 F4 ]* o; E8 r! a6 d
%利用原数据对BP网络仿真
an=sim(net,pn);
* y, J$ `6 d/ n, V& I  a8 n0 Wa=postmnmx(an,mint,maxt);
: L# X& S, K& |6 U* `
- d0 T& B- E% B1 F' T! j  S# A- h%仿真结果与原数据对比测试
$ p* h/ K4 M$ K# V' Px=1990:2009;
+ H% h$ N3 n" K9 W! Znewk=a(1,;
newh=a(2,;
/ x  f5 C* b- b$ \6 i3 Kfigure(2);
: b. Y$ ]) @! O) ]9 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');
7 W3 g0 H3 T' M$ G. K1 J# txlabel('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--+');
; R6 s# B/ |. e, t; T9 E1 C, [legend('网络输出货运量','实际货运量');
1 B7 A  P5 L/ p; H( Q; Xxlabel('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');
; G7 q, I" x! p, ^  J( q2 C+ n) o
/ w$ \3 H7 y/ E%新数据仿真
' n' s& n$ |6 J  z0 [pnew=[73.39,75.55
$ L5 q, J0 i- Y# v) T+ O0 f  V    3.9635,4.0975
    0.9880,1.0268];
pnewn=tramnmx(pnew,minp,maxp);
anewn=sim(net,pnewn);
anew=postmnmx(anewn,mint,maxt)

6 {3 W6 I  Y" g* n  Z( W1 e修改后程序为：
& H7 _1 B6 R5 f+ m- @- d" x%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];
2 g! h; c8 I8 Q4 L$ M& ]4 ysqjdcs=[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,...
( f! y! h  ~. H9 w2 d    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,...
# G* a1 M- b" A7 r8 y    0.56,0.59,0.59,0.67,0.69,0.79];
! @2 `- z% k2 }6 f. z2 @glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
; H  q1 X' d4 i. S; R' H( X    22598,25107,33442,36836,40548,42927,43462];
# j4 Y4 h7 p. G0 f  h6 eglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
3 B3 s2 O8 K) v+ N! m# n    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

, U: B4 |9 F- @4 Y5 @2 o%数据归一化
1 T) c& b. G; [* M7 r' [. |[pn,ps1]=mapminmax(p);
3 s3 v  W1 D8 B7 r7 y% M[tn,ps2]=mapminmax(t);

' j) ^$ a. O! T( ~4 y7 w2 v# s- G7 o8 v%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
+ C' k) t& X, L- s. ?0 _: {net.trainParam.show=1000;
5 d. S9 U$ \' F2 n8 z# G& m( ?net.trainParam.Lr=0.05;
! R  k% E( p1 `1 |- u' h3 Mnet.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);
/ z% \+ r0 W- p2 ^: ?: `4 V( va=mapminmax('reverse',an,ps2);
$ g' W, g! g, q- n$ I$ C! S
1 r( z1 O6 W( t! n) [) g2 U%仿真结果与原数据对比测试
x=1990:2009;
0 V7 k; ~7 G+ T* h1 dnewk=a(1,;
newh=a(2,;
$ }7 g) ?% S1 nfigure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
9 c; @1 s  j% o' o" D1 Clegend('网络输出客运量','实际客运量');
3 x/ v( q: a' cxlabel('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');
& p1 @4 g' A- b' f# d; ]! Psubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
: k8 N/ `+ @0 u6 o! q! ^0 S6 Tlegend('网络输出货运量','实际货运量');
xlabel('年份');ylabel('货运量、万吨');
! ]2 u# b& q( U* utitle('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
" X! m7 G* l7 @    3.9635,4.0975
    0.9880,1.0268];
8 S, m# O# P, I0 K) ~pnewn=mapminmax('apply',pnew,ps1);
2 j/ ], {! l  ^& Banewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
( M* Z0 K6 Y. _+ x3 N( y5 V（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. clc3 c- |7 M' O) z1 q6 s
   
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,...
   ( h$ ~. _0 C& q: k
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];4 o9 ^9 ^8 C8 r/ o9 K
   
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,...' E+ G5 B) V$ l% W9 a7 A; X
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   / M\" d* W$ y, W
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,.... @0 |0 X. }/ Q) g* e& `
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   7 r: o4 x3 V/ }5 l$ h
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   4 u9 e( c7 ~8 ]! c
9.      22598,25107,33442,36836,40548,42927,43462];  B* r: u5 E0 ^1 n7 A$ K# O9 w
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    , ]# c9 |$ Z3 s) D% O  J' u* B
11.      13320,16762,18673,20724,20803,21804];
    ' \- }: ~9 X: g! a% ?) R
12. p=[sqrs;sqjdcs;sqglmj];
    . k7 O( p- H' H: l
13. t=[glkyl;glhyl];
    9 t& l, M2 `: z# A% ^
14. %数据归一化( ?- Y% @) x& b6 Z
    
15. [pn,ps1]=mapminmax(p);
    9 h2 ~; Y- D7 P
16. [tn,ps2]=mapminmax(t);  f( l+ K- s5 f+ [* |
    
17. %dx=[-1,1;-1,1;-1,1];
    $ j/ r% ]  s, p. w
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]
    / y- k( O5 N1 n
19. 6 x- @  r9 v6 _0 W$ U6 @- e0 s! u
    
20. %BP网络训练; o* ]( ^  c\" n5 y! `6 G0 n2 @
    
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]
    3 W8 A  y9 Y/ I$ d( t0 ]
22. net.trainParam.show=1000;
    # w8 I* x- m& E8 C
23. net.trainParam.Lr=0.05;
    8 k8 [: E$ K) w3 O
24. net.trainParam.epochs=50000;
    \" g0 l8 E: O, v) Q
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]. S: V5 a+ r7 l! N+ a2 o. y8 m' u
    
26. net=train(net,pn,tn);4 n! y4 P: A2 b\" @7 S4 ~
    
27. 5 Y  Q4 {1 m* m; h5 ?! g- |+ a- k
    
28. %利用原数据对BP网络仿真
    ) E4 M2 d- ?; u% T0 x1 u6 f
29. an=sim(net,pn);
    , N0 }+ I$ u3 i5 \+ K2 v- n7 q
30. a=mapminmax('reverse',an,ps2);
    * k3 i1 H- X9 _& t3 z$ f) R' ]' n
31. 
    7 }2 B; s* L; G2 l! m' C/ a  ?
32. %仿真结果与原数据对比测试
    - J* q# |5 n- j& F3 n- b5 n5 U
33. x=1990:2009;& s0 z/ Y1 ~% f4 x6 E% b2 c5 o! a
    
34. newk=a(1,:);. K: ^6 c  M! x4 c; M- W# {) C
    
35. newh=a(2,:);0 E6 L% S8 ~/ y$ e/ Z  R2 }
    
36. figure(2);
    $ S$ K* J# \; x& Y\" Y2 m
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');( y1 @: X\" a* C* i
    
38. legend('网络输出客运量','实际客运量');
    / N1 n0 K' B% \8 f
39. xlabel('年份');ylabel('货运量、万人');
    4 `6 ?\" N, B) Q- \7 V& g/ _: i
40. title('运用工具箱客运量学习和测试对比图');
    # K# W/ L2 A+ N$ {+ J/ [
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    3 K  \8 ]7 {9 T% N2 d5 w7 Q
42. legend('网络输出货运量','实际货运量');7 t9 E. L  {! H: A
    
43. xlabel('年份');ylabel('货运量、万吨');7 K' N- o8 ^8 S( S2 a* o3 G# K
    
44. title('运用工具箱货运量学习和测试对比图');
    2 x0 V: I. b; @: F) X
45. 
    & P1 ^: P- B  a/ I
46. %新数据仿真+ H$ |5 `1 H) \. U
    
47. pnew=[73.39,75.55
    % f5 y6 Y1 G, G/ m7 @\" W  X
48.      3.9635,4.0975
    9 X: e7 T+ L$ X! o& k
49.      0.9880,1.0268];
    2 z, [% n5 {. N3 [
50. pnewn=mapminmax('apply',pnew,ps1);
    $ i+ M+ Z6 J4 D0 n( T
51. anewn=sim(net,pnewn);* E: F- A  W3 K1 U9 m* o6 _
    
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
   - v/ }; D3 M! I1 h# U! c# m- w
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,...
   1 q9 m6 }  Z: u. {6 K
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];' x5 u9 v, w9 e- q* ], L' 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,...
   ; p, B% c$ a, B* [+ [
5.     2.5,2.6,2.7,2.85,2.95,3.10];4 Y; p$ J. U& p2 t. 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,...* p+ T0 ~- D( s) r\" n
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];! A! A& g4 j3 r
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...$ b3 z% r+ @2 Y- @% j
   
9.      22598,25107,33442,36836,40548,42927,43462];
   & p& ~\" o: L; G- a
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    / D, |2 A( t: t9 r- t( H
11.      13320,16762,18673,20724,20803,21804];0 C8 b0 X. \! F7 Z: Y5 R& A! A
    
12. p=[sqrs;sqjdcs;sqglmj];- a9 G8 j/ M* C) N
    
13. t=[glkyl;glhyl];
    : o\" I1 s+ P' _' q
14. %数据归一化, \6 P7 O/ y3 V2 u
    
15. [pn,ps1]=mapminmax(p);
    , X/ u! V! r) f6 d% K; Q
16. [tn,ps2]=mapminmax(t);
    7 {- G' o3 D0 k  z/ L% ?+ p$ |
17. %dx=[-1,1;-1,1;-1,1];5 ]1 h0 `: l3 y/ n3 g  `; B4 f, Q4 p
    
18. p0=minmax(pn);t0=minmax(tn);
    ( M8 y& s3 j/ O# ?; e
19. 3 S+ y! _7 Y! n1 {\" e6 D
    
20. %BP网络训练5 Z7 M  H) }% z, [$ c% a
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
    + l$ g6 r* \# L% [9 D: G7 C& Q/ r
22. net.trainParam.show=1000;
    . \8 d, G/ z$ O* @2 h. F/ P, P/ C
23. net.trainParam.Lr=0.05;
    9 ^9 D# l4 X! h  K/ z6 Q
24. net.trainParam.epochs=50000;\" ?7 Q# i# `, X+ o$ W) B
    
25. net.trainParam.goal=0.65*10^(-5);
    / Y7 A+ K  I+ o, Y, J\" `8 w
26. net=train(net,pn,tn);: ?7 R  w& N( u) x) A
    
27. - a/ T% B7 F( s- ?2 X! J
    
28. %利用原数据对BP网络仿真
    7 g& t& R% Z, I; I1 ]& _3 ]
29. an=sim(net,pn);
    ! q8 c( x+ s\" x3 m1 o
30. a=mapminmax('reverse',an,ps2);
    ; g5 C5 j, G8 U% K( D0 b6 I0 z
31. ! V8 x- u$ w( {. v\" B+ W
    
32. %仿真结果与原数据对比测试
    . ~0 v/ N6 t9 j
33. x=1990:2009;
    ; N2 f: u2 v+ u
34. newk=a(1,:);
    2 t, E4 T3 [* F
35. newh=a(2,:);
    7 @# [- O8 [  ?) w
36. figure(2);
    9 d$ O/ g5 ?$ _! a1 p. E
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');# o  W2 X, J5 S9 y7 [
    
38. legend('网络输出客运量','实际客运量');8 I+ B6 ~( g; V\" ~8 h- N, h+ A3 d
    
39. xlabel('年份');ylabel('货运量、万人');5 L) A1 s* z# J- e0 f- ~8 ^( ~; a
    
40. title('运用工具箱客运量学习和测试对比图');$ F9 y$ Y- z0 l& i/ b$ o8 n
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');; y( ]' c# o2 c# V1 C2 [1 n* T
    
42. legend('网络输出货运量','实际货运量');8 ]* `6 g5 Y. T1 m/ o; q2 m( ~
    
43. xlabel('年份');ylabel('货运量、万吨');
    9 _7 f3 Y6 x/ f# W
44. title('运用工具箱货运量学习和测试对比图');$ Q+ I- c% K3 H, u* P
    
45. 
    * a1 l4 E6 H! D5 Y& G
46. %新数据仿真
    \" t& ]( C& n# K% D9 P0 f+ K& R
47. pnew=[73.39,75.55
    / |: j6 x: @. Q+ ^0 m
48.      3.9635,4.0975
    ; R/ Y& c5 Q7 \2 G
49.      0.9880,1.0268];
    9 D3 |6 ]8 X' q6 E; ]! c
50. pnewn=mapminmax('apply',pnew,ps1);
    ' F. P# m0 I1 r5 C* f
51. anewn=sim(net,pnewn);
    ; a* d6 |$ B) {, g
52. anew=mapminmax('reverse',anewn,ps2)

复制代码

T-Eric

且生 发表于 2014-7-21 18:22
5 m' j; G6 @: X+ W9 E忽略我上面那个，改动的地方在18，21，25行……求批

- z( I# H/ Y/ @3 e& D" J' x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
   ' z4 ?+ [; G7 G8 b6 r/ t- P: {
2. net=newff(p0,t0,6,{'tansig'},'traingd'); $ ]) M$ v5 ^3 J8 Z
   
3. net.trainParam.show=1000;  D$ H3 k& h. P- [( H4 o
   
4. net.trainParam.Lr=0.05;
   ; s( @  d- `( I7 ?\" R
5. net.trainParam.epochs=2000;
   6 i! u6 w; w' l0 @: s- J& H
6. net.trainParam.goal=0.65*10^(-4);1 N% A* |0 m) j+ h
   
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
- }" j0 c! n4 P: |. w1 R" I三、迭代次数改成2000
' t: G( b( {6 q' ^# B上面的参数是自己试的，我也不知道为什么。
' N4 `: r" K0 t% f" ?6 G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