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！！
9 H8 ^( ]% W- o2 F
%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,...
" X4 E" A' B1 J, R% h- ]    2.5,2.6,2.7,2.85,2.95,3.10];
0 U0 J: l  n' t! Z2 Ssqglmj=[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,...
2 }) S; s! k; b* a- n    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];
9 M" v% [& H) g3 |1 \3 M8 S8 o3 Vglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    13320,16762,18673,20724,20803,21804];
& T' k1 a6 ]+ u7 J: ?9 Z, {p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

2 R8 S  _/ o" l%数据归一化
[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
dx=[-1,1;-1,1;-1,1];
: p* p/ T4 {: L5 \
' E  i7 T/ X0 T2 w+ D* ?  S%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
- F) [! J( h: Y7 w& z" D  znet.trainParam.Lr=0.05;
8 \6 ?6 Z' f7 s) E; [net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
1 z; |* E; m1 d, z- q  mnet=train(net,pn,tn);
+ A  W7 \* C7 w( A( V; l. Q  }
+ {( ^; g4 t8 o) W/ S: z, p; A%利用原数据对BP网络仿真
' }1 G1 T6 v( Gan=sim(net,pn);
a=mapminmax('reverse',an,ps2);

" ^8 j% H1 u, D: u7 I9 x%仿真结果与原数据对比测试
3 U5 c# D3 E3 A# B6 Mx=1990:2009;
1 P- k& [+ g: t+ Rnewk=a(1,:);
newh=a(2,:);
+ e8 P0 _& f4 n( C" ~  Ffigure(2);
: S- s. W5 Y2 O# {9 N2 C$ esubplot(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--+');
  X7 L9 z( A; T0 e) f& Plegend('网络输出货运量','实际货运量');
xlabel('年份');ylabel('货运量、万吨');
8 x8 K! D) _+ M$ S% t; Htitle('运用工具箱货运量学习和测试对比图');

0 P5 {" E6 q  q% q%新数据仿真
1 [! R/ Q9 m9 c# Cpnew=[73.39,75.55
+ w* a6 g1 x& Y" @$ k; i    3.9635,4.0975
* f0 ^6 L; \: [7 n( W, |    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
2 a% O3 M# v/ [! B4 Janewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)

  h( O+ q2 v, i

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：
# @2 I( f) A: J+ y9 Znet=newff(dx,[3,7,2]{'tansig','tansig','purelin'},'traingdx');
我修改为：
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：
4 O1 s8 N% P  o3 R4 F5 U. d. }[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
/ G) s: Z" C( w; u' wa=postmnmx(an,mint,maxt);
# L: ]% m% C% l; n+ tpnewn=tramnmx(pnew,minp,maxp);
anew=postmnmx(anewn,mint,maxt)
我修改为：
# Y. e3 U- H" Z6 u[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
a=mapminmax('reverse',an,ps2);
pnewn=mapminmax('apply',pnew,ps1);
$ g% z& o" l2 l& ^anew=mapminmax('reverse',anewn,ps2)

4 T, W2 D9 G# }+ \$ P! Z4 w原程序为：
8 I; _5 c% u8 W%原始数据输入
' V  G% c: p. B) Y& T/ hclc
) [% K. z9 n, \2 Nsqrs=[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,...
# [# E& e6 V0 j6 r: }' J! A' }2 d    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];
; q$ @2 Q/ u5 V2 K" `9 Ep=[sqrs;sqjdcs;sqglmj];
+ l9 ~+ m  t( u$ B$ Y+ Z, f" c. d8 tt=[glkyl;glhyl];
) k, p$ f" U% B& D2 r- Q
%数据归一化
' W9 X/ N9 b2 l( g6 }' s% |& K[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
4 `6 C/ x$ Z+ z  v1 Xdx=[-1,1;-1,1;-1,1];

1 A3 G- R# L2 E& b& j%BP网络训练
: z( E$ o% O+ c0 _net=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
' Q# s. o: G$ K0 S4 e& anet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
9 j" c- w% R' L5 x. r5 n9 Q" cnet.trainParam.goal=0.65*10^(-3);
9 j/ F, U- y! v, t" I% Lnet=train(net,pn,tn);
7 @3 \! R# J1 c. p7 u7 D
! V1 M' U$ `+ }+ d$ z%利用原数据对BP网络仿真
, d6 n$ ^* {. H$ f. San=sim(net,pn);
a=postmnmx(an,mint,maxt);

%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,;
; B/ w" n* p: Inewh=a(2,;
" J1 \+ o  l; Y. J8 d7 v% _figure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
. n$ V* j# C: C& J7 v  U8 j9 elegend('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');
7 M) `2 D0 O: ]6 e  P7 R* u" ysubplot(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');
! `2 }8 V* a2 I/ m+ }% z$ ]title('运用工具箱货运量学习和测试对比图');
* H4 @* U! A# f  t
- u: S3 Z5 _2 c( t2 ?8 F%新数据仿真
pnew=[73.39,75.55
    3.9635,4.0975
$ x. G' |- G, ~    0.9880,1.0268];
, F1 b) ]$ c" K9 |9 Z7 \+ S9 upnewn=tramnmx(pnew,minp,maxp);
8 d/ ?  }% p0 nanewn=sim(net,pnewn);
+ j! S  n6 x! _' b$ y/ J9 Tanew=postmnmx(anewn,mint,maxt)
6 w& l2 g0 t- P8 Q' u. a
修改后程序为：
0 D5 W$ o) Z# s4 L& @4 X0 V2 }%原始数据输入
clc
: |- B# c: J2 ^# I+ V* rsqrs=[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];
5 t( x5 p* x  Q8 Z9 m& Vsqglmj=[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,...
, J# F" A* B2 R. Z$ p0 `" m    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];
3 N! O8 n! l/ s% N1 d4 c
* M- P" f' Q" E. m+ ~4 F1 K; q2 v%数据归一化
[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
4 X$ e7 R& f$ S) ?
%BP网络训练
4 R" M8 D% X5 a0 g, w$ A+ Qnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
net.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
8 ]( B! X/ m- d. o3 h1 w. xnet=train(net,pn,tn);
2 {+ `. x& }, w% y" f* l: @
%利用原数据对BP网络仿真
8 M3 Y' g- Z* _an=sim(net,pn);
0 k3 F; h0 I0 Z5 E) j, Ma=mapminmax('reverse',an,ps2);

%仿真结果与原数据对比测试
x=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');
1 f2 t- y0 D* R) R  oxlabel('年份');ylabel('货运量、万人');
' H/ P9 _7 ]$ wtitle('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--+');
6 K6 H. D% n7 H$ n/ q3 T$ blegend('网络输出货运量','实际货运量');
( u; S; }( n4 v# C+ Jxlabel('年份');ylabel('货运量、万吨');
1 J, \+ v: w! wtitle('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
% i% v& H4 v( v$ [. U( X; o5 P- h    0.9880,1.0268];
( _- C8 S) D2 A; Z) @pnewn=mapminmax('apply',pnew,ps1);
. F4 ?+ {2 F- F( Kanewn=sim(net,pnewn);
% ?6 H6 @) A) J3 I$ D- manew=mapminmax('reverse',anewn,ps2)
1 H6 N. ]: @, X, G（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. clc& g' B$ @1 `% ]) {- p3 L
   
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,...\" D% l+ t* X+ p) M. U2 H\" y! [; R
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   . M5 i; R4 J6 G7 E# w6 ?* S7 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,...
   $ a4 o: u% L! \\" f  y/ E
5.     2.5,2.6,2.7,2.85,2.95,3.10];# V% s1 {7 G6 U( a. u- g1 ?
   
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,...& z- w2 g% p- \% G( ^
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];' G4 ]2 K) j4 K0 A1 f2 x! y
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...$ z! ?$ f* H7 a* [4 D
   
9.      22598,25107,33442,36836,40548,42927,43462];8 {; d& d5 T# h, N8 r1 ]+ D
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...9 R/ \5 V- E2 }+ ?% r) ]\" K7 p. S
    
11.      13320,16762,18673,20724,20803,21804];
    \" M) ?! q  N7 j  Y2 n8 E5 m
12. p=[sqrs;sqjdcs;sqglmj];
    \" i$ s& c4 w+ v. m. O2 W
13. t=[glkyl;glhyl];' [7 {0 v. _: {; H# F, f% w
    
14. %数据归一化+ \9 Z) Z* t3 W% m2 g
    
15. [pn,ps1]=mapminmax(p);
    \" l# k2 K' @* G' u\" @% e
16. [tn,ps2]=mapminmax(t);5 k* {5 J+ }1 z4 S, C4 V
    
17. %dx=[-1,1;-1,1;-1,1];
    $ r' ~  s1 f8 v% w6 D' q9 h, _9 u  R
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]
    9 t' i) [/ U9 B
19. 6 @2 I: `$ J1 t: i6 F7 e
    
20. %BP网络训练8 Y! R: I\" d: l
    
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]
    \" u5 M/ n) N) ]) C7 k
22. net.trainParam.show=1000;* U  ~- t2 H1 e  u! g
    
23. net.trainParam.Lr=0.05;- y- K3 x  h  n2 |* ^
    
24. net.trainParam.epochs=50000;8 A2 u. {2 x1 I
    
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]
    * _, W9 U9 C7 {+ f* B% T! M
26. net=train(net,pn,tn);: l3 @# c$ ?: w' ~4 f# n: u% j
    
27. 
    ! r: x  P6 b6 Z2 K/ E$ s; s2 U2 k
28. %利用原数据对BP网络仿真! C. \! \\" a! b, o8 B/ ]0 P
    
29. an=sim(net,pn);, |+ z& H/ y3 ~8 p6 I6 t
    
30. a=mapminmax('reverse',an,ps2);& Q. |2 u5 y* \1 e. l
    
31. 
    - g  M6 {  Q( ]( s, L
32. %仿真结果与原数据对比测试
    % c8 O1 p' C1 T\" N' `! f- V
33. x=1990:2009;
    + P6 z4 a& i! a1 G1 T6 u# v
34. newk=a(1,:);
    $ C5 w# k\" G; w
35. newh=a(2,:);- ~# t' @5 }* A) A. u& T! I\" }/ v  Z) ?
    
36. figure(2);% [8 u( k1 k  T+ K7 V0 j1 Y; A
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    5 Z1 @, f  b* v3 D! O
38. legend('网络输出客运量','实际客运量');
    ; p: K* W0 C- K4 h
39. xlabel('年份');ylabel('货运量、万人');# D  Z# Q( t) @( Z7 X( d
    
40. title('运用工具箱客运量学习和测试对比图');) ?& P0 C2 u5 n: W; W
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');# n# G. X' O; `9 w: m# I$ X! Q
    
42. legend('网络输出货运量','实际货运量');6 {; F* e. l7 J  N  e# }1 O0 e
    
43. xlabel('年份');ylabel('货运量、万吨');$ q\" f) N, b6 a& R- X4 V2 R
    
44. title('运用工具箱货运量学习和测试对比图');; e5 `\" _5 U% ?7 q5 I
    
45. & n\" K( ]8 `7 e  {6 `* C, u
    
46. %新数据仿真6 |9 u1 \0 b  T! q; Z
    
47. pnew=[73.39,75.55
    ; M\" J2 X, H: I$ e* d
48.      3.9635,4.0975
    ! k( B  N7 e) g9 q8 O6 F
49.      0.9880,1.0268];* u' R' ]8 V! `( Y8 L3 x
    
50. pnewn=mapminmax('apply',pnew,ps1);
    6 [! i\" X8 o  m6 q
51. anewn=sim(net,pnewn);! W5 \\" z\" |' J, 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. clc; k+ _  r$ i+ ~( g+ ~% ?\" B# 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,...9 n) X) B( G+ @+ O
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];8 ~$ m/ a8 }: L  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,...2 m+ K2 o8 @7 c; O7 U9 G) b; R
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   : s- a; r7 o9 @, j& g6 W0 q
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,...# e6 d, u# {# f+ h
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];# \6 b# U3 n' ]4 }& b5 \
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   , `, u% z\" `& H$ \
9.      22598,25107,33442,36836,40548,42927,43462];
   . i+ ]6 g' i\" c% N  {
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,.../ o/ l; T! E8 I( W0 g+ n( y% r
    
11.      13320,16762,18673,20724,20803,21804];4 o5 @( }  q+ ?0 C7 y, O) i6 N
    
12. p=[sqrs;sqjdcs;sqglmj];
    ) d+ Z9 \! J3 g4 H. w8 G
13. t=[glkyl;glhyl];. m5 f! m1 L3 k1 G1 ?! n
    
14. %数据归一化7 w: u# R: d  j/ F
    
15. [pn,ps1]=mapminmax(p);! j+ p8 E3 \* @  e1 W7 k7 s
    
16. [tn,ps2]=mapminmax(t);
    / i* ~3 D; N. @
17. %dx=[-1,1;-1,1;-1,1];  `! x, X0 ?/ b- r9 Z' |4 n
    
18. p0=minmax(pn);t0=minmax(tn);
    + f: y) b0 t& C) \4 p) d
19. . o/ {4 k5 @, H\" A
    
20. %BP网络训练9 b8 }0 v; C. ^4 g2 v' e! n: n; x, \
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); 7 t9 h\" C/ |* K7 g9 Y1 w- D: J
    
22. net.trainParam.show=1000;2 L( P$ ~3 e! x4 k8 j
    
23. net.trainParam.Lr=0.05;6 s8 S7 T8 d+ A3 N7 _* ]
    
24. net.trainParam.epochs=50000;
    ( j  G* }/ l# D% I
25. net.trainParam.goal=0.65*10^(-5);
    , U5 n% f# x( b4 P% C$ @' P$ \
26. net=train(net,pn,tn);/ A, v, K( k) U  j. }2 N
    
27. * K9 v) T* B8 u& }8 ~1 J
    
28. %利用原数据对BP网络仿真% A3 ], U- c2 ?& B0 p) ?* ]
    
29. an=sim(net,pn);6 a* H3 K+ T5 A: o4 J1 i8 b) }- D
    
30. a=mapminmax('reverse',an,ps2);
    % C/ O9 t0 G' |1 a/ {
31. 
    + L* G* N( a2 y$ s; e1 o! N! K
32. %仿真结果与原数据对比测试 # y# ]  Q\" R! e2 m& o
    
33. x=1990:2009;
    & U\" q* r8 N7 U9 `4 C0 D& \8 Y
34. newk=a(1,:);
    1 a! c1 g# J% n9 W7 k3 b' g: P
35. newh=a(2,:);, E' n  K& i7 `# ~( ^\" N: b9 C7 b$ L
    
36. figure(2);1 \3 I7 }! ]0 S\" b
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    1 L# \8 k; D5 P, z& L0 [6 X
38. legend('网络输出客运量','实际客运量');0 k9 \8 v- `$ ?- |  f  ~: J
    
39. xlabel('年份');ylabel('货运量、万人');: @& j2 H1 P. z( m6 |
    
40. title('运用工具箱客运量学习和测试对比图');
    , Q2 d- ]# }' x9 n2 q, h
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    ! u; d9 j8 O\" j5 M6 S- l
42. legend('网络输出货运量','实际货运量');
    * g1 s; [7 s* z- I0 V. q2 D
43. xlabel('年份');ylabel('货运量、万吨');
    $ ~3 g$ Y4 _' o4 g
44. title('运用工具箱货运量学习和测试对比图');
    : i. _$ s% V5 {! g
45. # A\" s, b2 x9 f$ ^
    
46. %新数据仿真
    * S: M5 X) {/ c+ I8 X9 i
47. pnew=[73.39,75.55
    0 Q, H' {5 H+ r: y0 }3 b2 Y
48.      3.9635,4.0975
    2 A  X6 ?; x8 M: J
49.      0.9880,1.0268];6 i8 _6 l3 c; [7 ?# Q, _
    
50. pnewn=mapminmax('apply',pnew,ps1);
    7 z, C. g! q  c* g' X; {6 T& d
51. anewn=sim(net,pnewn);$ q: Y# L) I  Y5 @9 {$ ]
    
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

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\" n' ]/ @2 g7 {/ X2 u
   
2. net=newff(p0,t0,6,{'tansig'},'traingd');   A: q: C8 ?, w1 i0 z7 k7 {; i
   
3. net.trainParam.show=1000;
     B7 e* @( R) u7 n/ \
4. net.trainParam.Lr=0.05;2 ~$ {$ F( T' e\" c; B
   
5. net.trainParam.epochs=2000;  n% z+ a0 h, p7 O
   
6. net.trainParam.goal=0.65*10^(-4);
   # N1 A+ y0 M( d1 d! ]; U2 @' i
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
3 C3 y( x& d& b/ v" M  c+ o6 X二、训练函数改成梯度下降BP算法 traingd
6 Q+ |, y, Y) c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。
! u  K1 P' T2 Z3 t* t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