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！！
# L& O& A" T$ V; g' j( ^/ C
6 G  c+ t  M2 W5 L%原始数据输入
0 M- ], {7 H- `) e9 k" \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];
' D$ L# M* a9 L7 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,...
    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,...
  C2 ?4 T( }9 K, a    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];
* B8 L3 n- q8 g6 A" U2 P! b1 i, zglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
8 `0 c  Q1 U( b. Q. {    13320,16762,18673,20724,20803,21804];
1 ]. M/ O: o" y/ X6 Q7 d# y' Yp=[sqrs;sqjdcs;sqglmj];
6 e6 l+ \) z8 `, tt=[glkyl;glhyl];
6 m3 {6 j- m% @' S* H, d
%数据归一化
[pn,ps1]=mapminmax(p);
; X3 L  p& H  v0 {( C; s# x' j[tn,ps2]=mapminmax(t);
2 R4 e4 e8 D4 Ddx=[-1,1;-1,1;-1,1];
8 ~1 ^7 T7 K+ F  c, g. I, J
%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
, _# g0 v: c4 t+ \3 s, Hnet.trainParam.show=1000;
0 c- j3 a6 r, z/ C8 Ynet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
1 O$ C+ h- c# ^& P  R' q9 Xnet.trainParam.goal=0.65*10^(-3);
6 R' \4 V9 v" c1 n. \/ M- Ynet=train(net,pn,tn);

%利用原数据对BP网络仿真
an=sim(net,pn);
% F1 M* o/ h+ X: k: ua=mapminmax('reverse',an,ps2);

+ |$ d, I/ ?: z. l%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,:);
newh=a(2,:);
# e  d: [3 T7 L! }3 sfigure(2);
; q, @5 `! [2 X, \7 M! W) }% r! Fsubplot(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');
2 L: y' y3 A9 I9 Q( i  Mtitle('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--+');
+ ^) z. G8 ]7 I/ Z. B3 mlegend('网络输出货运量','实际货运量');
  g& a. y. d0 q% V+ @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');
( s' M4 k: ]5 |; _, J, b! @
& W. O) C  q7 R4 h%新数据仿真
, ^# o$ Z- d- e! ~" R4 fpnew=[73.39,75.55
    3.9635,4.0975
$ h% ~  Z5 K- S$ W& A    0.9880,1.0268];
, {: ~6 G" V: o3 Lpnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
0 ^6 O# b& q5 |' H

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 ...

5 Y, l8 b% `% K: [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：
: F4 n2 a4 X! k" O9 z. o1 Unet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
* b* n+ L- ]4 ~* c4 K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);
3 H2 v; m$ m5 M6 [4 ja=postmnmx(an,mint,maxt);
+ {& T# U+ b8 P  A0 p; q. apnewn=tramnmx(pnew,minp,maxp);
anew=postmnmx(anewn,mint,maxt)
我修改为：
+ D$ t! c0 R" \/ h. ~! }[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
a=mapminmax('reverse',an,ps2);
pnewn=mapminmax('apply',pnew,ps1);
2 e% b6 D# ^/ v' O+ Y3 g/ Sanew=mapminmax('reverse',anewn,ps2)
) u& I3 A$ p7 U; x
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,...
    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 m- S7 I/ e/ D3 T. d* g% \    0.56,0.59,0.59,0.67,0.69,0.79];
- E% d; ?2 `: G# a$ E7 f; k5 H; Fglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
7 Y) }7 e6 g/ Z: ]0 f3 |' n7 O    22598,25107,33442,36836,40548,42927,43462];
( J; J; M$ l* \" `  m) e6 i' `glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
% [( p, U- G2 n( f5 E. a2 d    13320,16762,18673,20724,20803,21804];
- X% }* v0 `; T4 H& _+ op=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];
! Z4 B* C( x2 k2 S4 ?! e. ]
%数据归一化
& Z! T+ b' D! D/ E. U[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
( O: Q% i' H5 p& Q- Fdx=[-1,1;-1,1;-1,1];
# n' }& p6 Z" Q3 Q7 }8 y
8 j( ~) P  r/ U%BP网络训练
net=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
) T! X+ u/ H. [0 R6 s" Gnet.trainParam.show=1000;
/ e6 h( V8 B+ nnet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
& R# ^1 x; E( f  g4 f  [net.trainParam.goal=0.65*10^(-3);
' @- Y: b  J# _net=train(net,pn,tn);
) B  |7 f) o" h$ I8 a
%利用原数据对BP网络仿真
5 u) b7 k, n/ N9 h6 Kan=sim(net,pn);
a=postmnmx(an,mint,maxt);
6 L2 f# y+ C. [) l! F0 Q( h- V# f
%仿真结果与原数据对比测试
$ D  `+ m* I2 M' \4 _x=1990:2009;
newk=a(1,;
newh=a(2,;
: |  O" q/ k* n' ]; {figure(2);
) `4 I5 I. {/ q( B# `0 J% ysubplot(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');
) u" X4 ^" P: m9 z3 {) a4 r$ dtitle('运用工具箱客运量学习和测试对比图');
) [* w& t; e  E. @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');
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');

8 c( ^7 j, f- s%新数据仿真
pnew=[73.39,75.55
% ^6 [6 M. B# t: z2 N1 D    3.9635,4.0975
8 W* o$ \! M9 `3 s3 z    0.9880,1.0268];
pnewn=tramnmx(pnew,minp,maxp);
anewn=sim(net,pnewn);
% z& a6 t, |$ hanew=postmnmx(anewn,mint,maxt)

1 h9 j* q* N0 C5 b5 c, y修改后程序为：
6 V6 I5 [7 j; a8 F. o%原始数据输入
- H; x! j3 S) w, t- H. X& c# Dclc
* w" K' T3 h/ ], v, {; y" K, tsqrs=[20.55,22.44,25.73,27.13,29.45,30.1,30.96,34.06,36.42,38.09,39.13,39.99,...
2 T! e; i  n. ]3 B' {$ q1 N. O6 q    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
) ]3 b) N" B4 J& P5 y! o7 o8 O9 wsqjdcs=[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];
% D' V: F# _/ _; K) Nsqglmj=[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,...
( \; S' B; q0 |* n0 K( Q# B, Z    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,...
7 K( m- R" L+ m, B% n+ k! p    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
( E0 S- p% w. f( kt=[glkyl;glhyl];
/ X; o  D# f4 Y2 s4 ?0 o+ u( n/ V
+ ?! b+ K. e- l" Y2 L7 [%数据归一化
  A& L+ t$ v+ T[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
) e1 l3 B' q) ^( k* N: e3 k. L
9 |! B7 H8 d3 O: T%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
; h! N. a7 y0 x: Tnet.trainParam.Lr=0.05;
8 J" |5 T; [: ?: k) k1 [net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
# C0 w2 _! _  h% P1 _) `
%利用原数据对BP网络仿真
an=sim(net,pn);
; }% l9 C0 ^3 Ia=mapminmax('reverse',an,ps2);
8 B; ~( K4 k; `, r( u( |3 l
, ^3 P9 ]9 k, n" x& l%仿真结果与原数据对比测试
( H- g5 \9 A0 X' G; T3 e9 Xx=1990:2009;
newk=a(1,;
2 e9 g/ ^$ q; \) m, ynewh=a(2,;
9 I4 {: ~5 ]4 Ufigure(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');
: J2 y8 u9 K# y' ]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');
+ d% W: U  Z! s3 v( pxlabel('年份');ylabel('货运量、万吨');
3 M4 ], k' \# N; H; atitle('运用工具箱货运量学习和测试对比图');

1 A# `: ^6 u& M3 ^9 s! P4 M+ o%新数据仿真
8 I  N- m. q) Z" \3 d6 Wpnew=[73.39,75.55
    3.9635,4.0975
# C, o8 m% ^+ b! L    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
（修改的地方用颜色标记了）
6 ?  K5 ?8 Z5 i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; N% n\" D9 @; b: p
   
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: {9 v& K% \- R# }7 e7 O
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   . W0 L/ X* v0 n1 E* p) b2 _7 h6 c
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,...
   3 y3 ?5 v: J6 Q8 e/ O6 a
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   , A; _! L6 ^& ?$ h7 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,...
   . K+ s2 y  C) {+ Q. I, H
7.     0.56,0.59,0.59,0.67,0.69,0.79];/ ?7 o5 i\" M5 V
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...  E& i- Q* p* `% v! m# ?
   
9.      22598,25107,33442,36836,40548,42927,43462];
   7 _1 E( q6 o: k0 W$ P& K7 N4 E
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...4 K\" S3 V$ P% G( z4 \
    
11.      13320,16762,18673,20724,20803,21804];
    / c1 C0 m& z8 J4 f# q, e7 H
12. p=[sqrs;sqjdcs;sqglmj];
    2 I5 i1 b: c% W* H6 W# U0 h
13. t=[glkyl;glhyl];
    7 r2 |/ L: `4 W% T9 o% K. `( |4 c  k
14. %数据归一化
    0 J7 p- m( Z( }, `2 i
15. [pn,ps1]=mapminmax(p);
    9 z4 B# \\" E\" n0 U; U
16. [tn,ps2]=mapminmax(t);
    5 ^. p\" s9 l6 a- D& s$ w
17. %dx=[-1,1;-1,1;-1,1];
    4 x* ^  d# b; [6 }# m2 ?( [- @+ f
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]6 {4 A' O/ ]% V
    
19. 
    ! V1 C7 C9 ~/ o6 R8 R7 B
20. %BP网络训练! |3 r  y) ]* }. ?+ g
    
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]
    ( k6 q1 K4 e9 w3 `0 A\" }
22. net.trainParam.show=1000;+ k  [8 f2 {& s5 N7 J0 m
    
23. net.trainParam.Lr=0.05;
      J6 f3 f4 {6 k
24. net.trainParam.epochs=50000;$ R8 y( Z) [/ n: C
    
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]: D7 y1 i6 l! w7 V$ Z6 N
    
26. net=train(net,pn,tn);+ b+ X5 R: ?! [
    
27. 
    9 X\" w4 U* r) ^9 F4 w
28. %利用原数据对BP网络仿真
    , p9 J' P$ C# w( F. r; p
29. an=sim(net,pn);
    7 x- n2 c) h\" [, i
30. a=mapminmax('reverse',an,ps2);
    : V, r! P9 n4 _& p4 S; Q  k, ^
31. 
    3 Q/ m/ P' {9 T3 G' c( \! j6 x. I
32. %仿真结果与原数据对比测试
    ( E' T0 g! z7 r) _7 z9 j1 a
33. x=1990:2009;) J9 P2 z( p! |. l: q$ f1 u, U1 J; }
    
34. newk=a(1,:);
    ! T. i! I2 g( Z, t2 m2 r; M4 h
35. newh=a(2,:);
    ; s( q. Y0 N) I/ _1 R
36. figure(2);7 v0 b1 U& \/ [4 s4 |. A6 z
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');9 `( h7 W7 G+ g8 j7 o- ^; o8 W6 o
    
38. legend('网络输出客运量','实际客运量');
    ) h: F/ l* P/ S: k
39. xlabel('年份');ylabel('货运量、万人');
    $ [\" T+ Z+ Y  r- |5 S6 ^4 v
40. title('运用工具箱客运量学习和测试对比图');$ E\" M/ D, a! w* G2 T
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');2 Z1 U, R9 u  }( l
    
42. legend('网络输出货运量','实际货运量');
    3 _\" }0 V( `) L) j0 j0 H
43. xlabel('年份');ylabel('货运量、万吨');
    8 \: w9 s4 k! c# c$ Q
44. title('运用工具箱货运量学习和测试对比图');
    . q6 ?' h% {\" I( }1 s# U
45. ( @' r! [- v3 a
    
46. %新数据仿真8 R9 H( [  x/ J
    
47. pnew=[73.39,75.55; F5 k8 _& Z: {: N4 T* R
    
48.      3.9635,4.0975' _  T0 N3 D/ w1 E$ Q# x$ K2 s
    
49.      0.9880,1.0268];+ L! @; v' t* e1 ^
    
50. pnewn=mapminmax('apply',pnew,ps1);) t, G3 `9 Z) X/ l- `, G
    
51. anewn=sim(net,pnewn);4 Q\" J* J! ^8 x0 n, Z
    
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
   7 ~7 U2 E- e: p; O0 F
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 o/ I% N# a$ I8 v7 G( W
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   8 u2 q0 b- [\" R3 h  b& e& l7 q
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,...( h: V+ M& _8 x1 _: ~
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];! p( r8 X3 h\" 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,...
   * ?) a9 X2 Z$ @0 l3 f
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   ' Q# Y6 q- U1 N: q/ y
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...+ {, v# |  p* G; ]
   
9.      22598,25107,33442,36836,40548,42927,43462];
   ( I' C3 C/ A& M& S
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    & Q4 q+ j  A$ X\" O
11.      13320,16762,18673,20724,20803,21804];
    # m  o  _; u8 k$ U+ f, m
12. p=[sqrs;sqjdcs;sqglmj];
    * R8 g; n. H1 u: F9 X& V  G
13. t=[glkyl;glhyl];- O+ ]1 E% G. ]$ U2 I5 I
    
14. %数据归一化. y7 A) e0 O) f0 f
    
15. [pn,ps1]=mapminmax(p);- a( l5 t# ?/ I( x
    
16. [tn,ps2]=mapminmax(t);) B3 g+ k7 f! W- C
    
17. %dx=[-1,1;-1,1;-1,1];
    \" h, x1 ~% f& b! W1 i% g
18. p0=minmax(pn);t0=minmax(tn);8 m8 G2 t5 X$ ~( {. s+ b! h; \7 f! u
    
19. / f( C, F9 U/ ^; |2 ^+ `
    
20. %BP网络训练9 W$ V$ U7 T, t2 w* Z# m* R! q8 m
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); ! T1 s# C$ W* {2 B4 Q5 P3 g% d5 [$ J
    
22. net.trainParam.show=1000;0 n1 b) R/ k4 ^! [
    
23. net.trainParam.Lr=0.05;$ F  B+ y  t) D( H% i( v# t
    
24. net.trainParam.epochs=50000;
    # i) U2 E8 G$ G
25. net.trainParam.goal=0.65*10^(-5);  ]( j, u. S7 K) B( t! k
    
26. net=train(net,pn,tn);( Z5 W) Z% ^! Y+ Q: G/ u. V' u
    
27. 
    ; {0 O( }0 k7 L) u, e* b
28. %利用原数据对BP网络仿真
    4 i% `3 z. m& c6 o
29. an=sim(net,pn);, k; _  Z/ j1 ]1 P
    
30. a=mapminmax('reverse',an,ps2);5 q. Y( E% |/ }5 j\" h2 {, [- B4 W
    
31. 3 I  ~* g8 V8 V! `! X8 x
    
32. %仿真结果与原数据对比测试
    : h+ h5 S5 o+ N8 j7 N9 B
33. x=1990:2009;# @* [: r6 F3 P+ L+ W! G- z\" l2 L
    
34. newk=a(1,:);( M5 \+ P' V7 X! n* k6 l9 z
    
35. newh=a(2,:);
    4 R, K4 W\" F( k* E. S
36. figure(2);0 }+ L: O7 d, g% P( ]% j
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    4 b3 F4 [\" e; L
38. legend('网络输出客运量','实际客运量');
    ' c6 P# ?' I3 P3 P4 G. }
39. xlabel('年份');ylabel('货运量、万人');# \4 N6 E3 }8 `\" U: Q' x
    
40. title('运用工具箱客运量学习和测试对比图');
    , Y6 u3 T5 c) e9 c
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    / R) `\" J0 u' m3 Q
42. legend('网络输出货运量','实际货运量');
    & H: Q, C( m\" {3 `6 R# e$ P
43. xlabel('年份');ylabel('货运量、万吨');
    0 `7 {$ Z$ R! P8 t0 d8 \+ D
44. title('运用工具箱货运量学习和测试对比图');
    2 {. S! ]5 Z! K3 [
45. - B( {3 `9 m4 p+ g2 }
    
46. %新数据仿真
    # B4 Q  t# h5 g/ w; N, f* e& M
47. pnew=[73.39,75.55  x- i9 s0 a, ]! Y  D# h\" t( v5 S
    
48.      3.9635,4.0975
    ; G- \8 G. a* N, W3 e$ ?$ L
49.      0.9880,1.0268];/ L  ]\" K\" a- t6 N* ^
    
50. pnewn=mapminmax('apply',pnew,ps1);
      m7 Z: N4 V3 F6 q% K8 W
51. anewn=sim(net,pnewn);( S4 F3 a6 F% N2 ]& y+ h2 W
    
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

4 I" X" P+ _9 A) q# g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3 \+ V1 B7 \% }\" u
   
2. net=newff(p0,t0,6,{'tansig'},'traingd');
   0 j\" I0 e1 z9 K1 y
3. net.trainParam.show=1000;
   : |6 r9 t) t, o7 W- a
4. net.trainParam.Lr=0.05;2 t5 b$ T( S/ K; {
   
5. net.trainParam.epochs=2000;. D( }1 T8 F7 r9 ^  x  X
   
6. net.trainParam.goal=0.65*10^(-4);- z% F2 V4 F$ ]: E- u' i7 \
   
7. net=train(net,pn,tn);

复制代码

我只改了这里面的，
& ~) {7 X2 G, w* L5 W9 ~- U3 x一、改成单隐含层的，6个节点
3 v2 t' `- J% Y, ]: W  Q$ ]3 v! n二、训练函数改成梯度下降BP算法 traingd
; {, K' i5 f% o+ r2 I- U三、迭代次数改成2000
! S$ W" g" l& [  {9 s' H上面的参数是自己试的，我也不知道为什么。
, a$ h* l) i& N: U' l6 ]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。
/ _! O1 P, l* G- |, k: E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