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* }3 C4 w; y' G  ^: Q1 V
4 ?2 O/ l. [. P! U0 F/ r) \%原始数据输入
& t5 D% c- v3 X6 Jclc
8 i, E/ i! G& o' r" `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,...
' c0 @' ~% {% 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,...
    2.5,2.6,2.7,2.85,2.95,3.10];
& r/ _5 k+ g8 j5 `6 h! r2 ysqglmj=[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];
6 r/ Y0 F& g$ d. i3 z5 xglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
. i$ {5 d: b8 a* s/ R    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];
6 |5 {0 u* s7 q0 Z9 u- l
%数据归一化
[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
dx=[-1,1;-1,1;-1,1];
2 t1 V6 w6 s4 c& N2 }6 _
%BP网络训练
8 p9 _1 u2 Z' r' l0 unet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
8 A9 k( T, |3 V3 N1 lnet.trainParam.show=1000;
; s, U' a8 h- D8 y6 C$ Knet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
( z* J5 Z( r& ]# D3 b& dnet=train(net,pn,tn);

%利用原数据对BP网络仿真
$ _* n; t6 t; o8 r4 pan=sim(net,pn);
a=mapminmax('reverse',an,ps2);

%仿真结果与原数据对比测试
6 P( c$ e  N- W1 c+ t/ jx=1990:2009;
. i' |* _( d+ Unewk=a(1,:);
* i( z* Z* s2 m* Z0 E( }newh=a(2,:);
figure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
5 i! m; c% D7 Ylegend('网络输出客运量','实际客运量');
4 P3 u1 k/ X  I& axlabel('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');
( O8 _1 |5 D2 `  Y; j3 \, m, _2 zsubplot(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');
' n0 b7 {, J! i* r; xtitle('运用工具箱货运量学习和测试对比图');

$ [# a. m% b$ S$ @%新数据仿真
pnew=[73.39,75.55
. j/ j  n4 F# ]; N    3.9635,4.0975
    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
2 W  @9 e$ Q" f( ?" T! V
7 i5 |( g  r# S' R, d( H, l8 Q

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
4 H# I/ P+ ^/ b, x! v- J8 I0 {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 ...

9 y, q$ W! j  ^4 r2 h. `因为版本更新，Matlab中的Newff命令用法有所改变，原命令为：
; Z; {. J/ T  o' _# `& G3 Rnet=newff(dx,[3,7,2]{'tansig','tansig','purelin'},'traingdx');
" |; K/ F- D1 V6 V我修改为：
; _* U( K8 X8 \( P/ f. Ynet=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：
/ g) G1 N" I3 l/ p[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
+ s$ e! w1 x  `$ wa=postmnmx(an,mint,maxt);
) R, y. k" {$ `pnewn=tramnmx(pnew,minp,maxp);
anew=postmnmx(anewn,mint,maxt)
7 \5 @5 o6 V9 i7 K: W9 n我修改为：
[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
a=mapminmax('reverse',an,ps2);
pnewn=mapminmax('apply',pnew,ps1);
  ^4 h/ p, j6 @6 Yanew=mapminmax('reverse',anewn,ps2)

: F0 {6 D7 Z+ z, E: \! l原程序为：
%原始数据输入
  l1 x7 G* k0 S: ?7 {clc
: i# g  \) A8 }. V, O1 usqrs=[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,...
    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 K" ?  |9 a! B. M" i* ^    0.56,0.59,0.59,0.67,0.69,0.79];
8 ]. N& s' e% K' y2 ]glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
! G9 o9 }+ _" R7 P, g- L    22598,25107,33442,36836,40548,42927,43462];
. ^7 C: }: C+ [7 o% E; E2 q. ]glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    13320,16762,18673,20724,20803,21804];
9 b, T$ t1 p8 ?5 E( k3 @/ R) h* hp=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];
$ B% g! F, B3 \7 J3 X+ L% f
5 ^& s+ @$ e( j/ ]( A' e%数据归一化
[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
/ z; N! I1 P# t7 Qdx=[-1,1;-1,1;-1,1];

%BP网络训练
. e# k0 [, g2 b1 `net=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
/ U8 e0 \7 b# B, D/ Xnet.trainParam.show=1000;
% L' S% ]2 V7 [" _) @+ ?net.trainParam.Lr=0.05;
4 u9 g7 W% P# n) \* w$ y/ H5 unet.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
3 S1 o. E; t0 r/ L# ?5 _net=train(net,pn,tn);
  ~1 L- i( Q6 p# K/ g% K" G6 K: O
/ C, G) [7 q8 s& o%利用原数据对BP网络仿真
/ s5 }' B! B2 z4 Pan=sim(net,pn);
6 Y: [6 n( ~# {5 ~a=postmnmx(an,mint,maxt);
6 f" V$ E+ b% I1 \4 w- y; |  Q
%仿真结果与原数据对比测试
x=1990:2009;
7 C) S2 t5 q4 L- d+ qnewk=a(1,;
newh=a(2,;
figure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
6 {) S* A7 V# ~- E: M) nlegend('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--+');
, ?1 \! d" c4 d' R( u/ Olegend('网络输出货运量','实际货运量');
3 C% v5 |- C# o1 xxlabel('年份');ylabel('货运量、万吨');
. w4 G* u. E: stitle('运用工具箱货运量学习和测试对比图');

+ i" b. F9 f, [' f- p6 j%新数据仿真
pnew=[73.39,75.55
$ `. l) `6 J4 o% V    3.9635,4.0975
+ u( X. H4 f1 |# F% A    0.9880,1.0268];
pnewn=tramnmx(pnew,minp,maxp);
1 s7 q- K4 t' Y2 uanewn=sim(net,pnewn);
9 _: k+ B1 R# a4 T. ^anew=postmnmx(anewn,mint,maxt)
' {9 d" `6 ?8 ^/ f  A, P
7 g9 F) q, n/ x+ p5 w修改后程序为：
  i7 b& x0 B2 h# h2 u%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,...
& U2 F8 g+ Q# I( K4 K8 U    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
% s% e, p2 ]+ ]# I  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,...
# T- |9 {9 R/ Z    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];
# ~; l/ M. V! C& Q! ?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,...
; H2 |7 b6 `# `3 s# ~! N: \& K5 w    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
& f: ?. F0 E+ G/ O( ~% Kt=[glkyl;glhyl];
. ^- G. g) N* Y
%数据归一化
[pn,ps1]=mapminmax(p);
4 b! k' X, t- d6 }[tn,ps2]=mapminmax(t);

  \6 Z% k6 J5 N# `' S, f, ^%BP网络训练
! @( q9 c2 t5 f) D4 M2 c) @net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
: ~. v2 g  O/ K5 h" Bnet.trainParam.show=1000;
# F' Z8 v9 V9 q/ j( Pnet.trainParam.Lr=0.05;
$ R+ P. c7 N! ~. U5 X! i& T  ?net.trainParam.epochs=50000;
! v' @( Y! ?3 N" r; Dnet.trainParam.goal=0.65*10^(-3);
+ W) N" t& g; r. t) s( |" ]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);

%仿真结果与原数据对比测试
- e+ {# e$ G. X7 ]( zx=1990:2009;
: z, q; j  B" |& P, ], y' cnewk=a(1,;
7 x/ ~) V) O3 i1 o8 I+ a' vnewh=a(2,;
figure(2);
! G% H! E$ Z* c4 y) ^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');
$ p' e  l; G! k& Ssubplot(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');
+ C1 R5 S9 u7 ztitle('运用工具箱货运量学习和测试对比图');

; O2 v+ H, i: i$ |. }%新数据仿真
pnew=[73.39,75.55
3 {- a4 j: t1 v8 V    3.9635,4.0975
    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
2 S( S: ^* `8 T8 }$ O' q& H! \anewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
" G' j$ f" i, u（修改的地方用颜色标记了）
, m& a* O+ |' u3 i& r5 E6 `8 E5 X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  \4 E+ ~0 f6 ?# ^6 Y0 a
   
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,...; G' j# }/ u# U- {' {1 y0 ?2 @; c
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];\" y' c; W$ V4 S9 `% b8 l* u$ R
   
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 s# n6 J8 V\" l
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];8 }  J# L. B* I' E+ t
   
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,...' d* r! E! y1 }
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];  H& a: V5 F7 s
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   ' i4 }) u6 i# d
9.      22598,25107,33442,36836,40548,42927,43462];
   8 A: C+ [3 e* o\" ]5 V
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    , H* `8 u- K! G7 G
11.      13320,16762,18673,20724,20803,21804];
    4 n. K# d' f) J1 Z( p
12. p=[sqrs;sqjdcs;sqglmj];& V$ X; e& q8 c* L: W% u0 N
    
13. t=[glkyl;glhyl];; R, g) {) y/ @9 v1 |
    
14. %数据归一化' z7 K9 V  W9 z- b% l
    
15. [pn,ps1]=mapminmax(p);
    9 `0 P' I$ |. u1 ~9 i6 w4 B, _
16. [tn,ps2]=mapminmax(t);& t  v: y% O- H5 S- F5 g) u
    
17. %dx=[-1,1;-1,1;-1,1];
    # c$ B) y  f0 U\" E0 K
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]
    , r( x8 T1 r- ~
19. 
      z! `, d. ^# U/ _1 Q5 l
20. %BP网络训练% R8 X7 e* [5 n4 {
    
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]
    ! P- \- J, @  j+ P: @9 i
22. net.trainParam.show=1000;
    / n( n& R! l( Q$ |) J* |2 U
23. net.trainParam.Lr=0.05;
    # c0 u! w2 G: l/ A9 u
24. net.trainParam.epochs=50000;
    4 s. Q, l2 }9 I, Y9 P7 j: {
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]2 ~1 B4 z' G/ e\" ]$ j+ _: q5 O0 t
    
26. net=train(net,pn,tn);
    * \& _& `$ P' T# n0 _; Q- y# c
27. 
    0 \2 [% h5 O8 \$ G0 Q2 E
28. %利用原数据对BP网络仿真
    ! T2 m2 o3 j, c5 V
29. an=sim(net,pn);
    ( ?* b. q0 C' p( y2 e, [7 ]
30. a=mapminmax('reverse',an,ps2);# O6 B4 Q, N9 K0 r4 Q8 p
    
31. 5 i. R; \# U# D3 F, _
    
32. %仿真结果与原数据对比测试 3 a+ E% M$ v5 d\" S0 _\" z. v0 e1 Y1 b
    
33. x=1990:2009;
    8 J3 ?! P6 n& G: e3 h5 a
34. newk=a(1,:);
    5 {: k& U9 Q: {! i! ?( l& f\" {
35. newh=a(2,:);! M( |3 x# R2 g* X8 w, M
    
36. figure(2);
    # p6 G9 N& S0 @. c  U) G6 B
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    ) d2 n# m  g& s! G
38. legend('网络输出客运量','实际客运量');+ J) R\" a! I1 b2 F
    
39. xlabel('年份');ylabel('货运量、万人');9 M% N8 W2 S2 p* B5 j) ]
    
40. title('运用工具箱客运量学习和测试对比图');- |3 A' s9 _$ ~) x! ^\" E/ J
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');+ @- |1 ^1 `6 n' U* X2 b
    
42. legend('网络输出货运量','实际货运量');
    / Q) f% M+ I$ @
43. xlabel('年份');ylabel('货运量、万吨');+ a# c/ P7 \% C' E3 G
    
44. title('运用工具箱货运量学习和测试对比图');4 q2 J/ G4 E- D
    
45. % N+ L# I& c3 f% _6 P
    
46. %新数据仿真; C3 Z( O% x% R$ o4 T* m
    
47. pnew=[73.39,75.55
    1 C' _1 U1 f9 ^# |5 m& N
48.      3.9635,4.0975
    % V& v7 o0 S4 X/ N. F- ^& v; Z
49.      0.9880,1.0268];7 i8 \\" C5 s9 W# s5 r2 H8 o, {
    
50. pnewn=mapminmax('apply',pnew,ps1);5 S3 u/ H7 f+ k
    
51. anewn=sim(net,pnewn);0 _( B+ {% L; M; @5 s6 x
    
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. clc7 {5 G0 q1 P( Q
   
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& O- B' r\" r! G; _
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   9 n6 b  Q+ e8 @. ?9 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,...
   8 w. K8 ^9 \9 S; o  A) {! u
5.     2.5,2.6,2.7,2.85,2.95,3.10];# N2 [1 P8 [' L: Z8 I8 @8 F( F* o! 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,...
   + y3 J4 K1 v7 x* L5 n2 p& [0 y
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   + y6 p3 y* B, r+ x) Z+ @% T
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
     M4 b# P, J# `
9.      22598,25107,33442,36836,40548,42927,43462];) U+ B& M9 D5 G
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    7 H! a' w: n. A# x$ x! q8 ~
11.      13320,16762,18673,20724,20803,21804];
    : v- H3 o* K' ]1 x$ F
12. p=[sqrs;sqjdcs;sqglmj];
    5 I* O) G, y; X8 l0 x( s$ k7 S
13. t=[glkyl;glhyl];
    : k1 O0 s1 C8 j# u% D  j0 n* m
14. %数据归一化$ V& q, j3 g# C
    
15. [pn,ps1]=mapminmax(p);% e( C6 e1 u! X/ x3 C
    
16. [tn,ps2]=mapminmax(t);% `- L. W& \% i
    
17. %dx=[-1,1;-1,1;-1,1];( l3 `# w! K1 ]* z6 R3 g- J
    
18. p0=minmax(pn);t0=minmax(tn);. O/ j0 M+ a! E: I1 N\" Z
    
19. 6 V8 n8 O( ^1 {, K; i# S# Z+ |
    
20. %BP网络训练1 i- o5 u# q0 ?\" p& t: z
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); 4 O! ~* [0 K! ^: J  ~0 k
    
22. net.trainParam.show=1000;  p/ N- S8 B4 e' n
    
23. net.trainParam.Lr=0.05;
    5 _\" @; A  n& f% N: e
24. net.trainParam.epochs=50000;
    0 J5 U9 \/ t- Q
25. net.trainParam.goal=0.65*10^(-5);1 M7 I) h. h- P! p2 [! F- K% l
    
26. net=train(net,pn,tn);
    # ~- R- y' q* z
27. 
    8 ^7 M, O, _4 M  K9 o3 ?& {3 c
28. %利用原数据对BP网络仿真# {0 v5 b4 A( t8 ^1 R1 C% L0 ^2 _
    
29. an=sim(net,pn);
    * k1 y7 A7 I! S4 J0 C
30. a=mapminmax('reverse',an,ps2);0 s+ Y- N4 Y7 M. X8 ]
    
31. 9 H9 g5 J+ X' M  L  W
    
32. %仿真结果与原数据对比测试 3 i4 D: a4 D. Y2 ?# j
    
33. x=1990:2009;; l/ g) R) }2 {, U
    
34. newk=a(1,:);
    + Y& e, D  A: j& k) ^% W
35. newh=a(2,:);$ [/ A; A, K. O- }( f
    
36. figure(2);
    2 v! s! m3 W' U5 E& V: @
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');% b. h* I\" b! l8 l& U; \
    
38. legend('网络输出客运量','实际客运量');
    1 ?# J+ y7 w. L: ~. U8 j/ ^' l
39. xlabel('年份');ylabel('货运量、万人');; a/ g/ w5 G2 k0 J( {4 F) U/ t9 @$ i
    
40. title('运用工具箱客运量学习和测试对比图');- t. h% p: }5 f; p  _
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    ) ~$ U\" x6 }5 y7 Z2 n$ H
42. legend('网络输出货运量','实际货运量');
    & s: l7 G/ i\" J$ k2 n9 a% ?7 j
43. xlabel('年份');ylabel('货运量、万吨');
    0 _. N- c8 D; w+ g, Z; J+ O
44. title('运用工具箱货运量学习和测试对比图');2 z3 W: `4 }0 ?$ u
    
45. & p0 c\" ^6 g' w; ?/ S1 J
    
46. %新数据仿真
    3 H8 k* l; f) X1 B9 R& K
47. pnew=[73.39,75.55. Y/ j0 t! L; C; J& h' a
    
48.      3.9635,4.0975/ c/ N1 a5 H; R% s2 T
    
49.      0.9880,1.0268];
    \" A) j4 G: }  g9 d7 m
50. pnewn=mapminmax('apply',pnew,ps1);
    & z4 Y3 Z! Q+ O7 m% W5 v
51. anewn=sim(net,pnewn);\" u3 n8 f3 K# H/ v
    
52. anew=mapminmax('reverse',anewn,ps2)

复制代码

T-Eric

且生 发表于 2014-7-21 18:22
- G( Y, ~# b2 p; ]% ^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
   ! @1 ~3 ~* _' G! z4 t) d
2. net=newff(p0,t0,6,{'tansig'},'traingd');
   # z2 R* {# O) h4 B! s
3. net.trainParam.show=1000;
   5 n\" z$ F* _6 k2 V$ p. Y8 W
4. net.trainParam.Lr=0.05;; D3 p: N\" t\" n4 o
   
5. net.trainParam.epochs=2000;) v  z% v% c) N6 i
   
6. net.trainParam.goal=0.65*10^(-4);$ G# k# G6 X\" _2 ]
   
7. net=train(net,pn,tn);

复制代码

我只改了这里面的，
: I: G2 H6 D! r! r8 W) H: [/ w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
. x, C0 G8 Q$ h$ d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