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！！
/ @/ w6 {+ a. {2 u9 ?+ v
5 k4 ^! F  K* J( i% y%原始数据输入
clc
  o/ r" O6 R5 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,...
/ O$ E9 x7 e# X$ [$ b1 U  S4 t' x    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,...
' z* \2 C1 w7 k    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,...
" M6 e% j8 O( o2 X; A: ]/ S0 A    22598,25107,33442,36836,40548,42927,43462];
8 p  R, s, e- _- m1 q! J) B5 r1 Jglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
- f  f/ V! u: F+ ]    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

%数据归一化
[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
  h; I" H2 g0 W/ Ddx=[-1,1;-1,1;-1,1];

% `: d" A7 i. q%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
/ a, x. H. T2 A  {0 [4 Fnet.trainParam.show=1000;
net.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
* g7 h* v3 P3 A
9 v( m& N5 Q( E5 R7 n  v%利用原数据对BP网络仿真
an=sim(net,pn);
a=mapminmax('reverse',an,ps2);
: }" m" Z" f" V% U
! H  C1 o' Q9 N5 d%仿真结果与原数据对比测试
x=1990:2009;
' C; f7 w; U2 f9 v3 T8 Gnewk=a(1,:);
' ?. J5 A+ U+ @2 Jnewh=a(2,:);
. }1 z7 Y* ]$ f; @: afigure(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');
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');
6 t  r/ c$ L' _  [
, N: `$ O2 T9 j9 o/ A) u9 R" r4 Y%新数据仿真
& K1 E4 L+ F9 x$ Ypnew=[73.39,75.55
9 _  i: G8 N+ {2 ~$ G& |+ Q    3.9635,4.0975
( g- G* ]9 \8 O; F9 R% {' X    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
' r. q9 V7 \0 ]6 a. c+ tanew=mapminmax('reverse',anewn,ps2)

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
, {) L& \8 R4 p) Q. t6 b5 t  y5 ^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 ...

$ V: a& R; A& D( ^2 \1 s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');
# j% e. `, @$ q5 K/ @6 s我修改为：
% L# j/ S+ Q$ b% j5 Jnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
+ s1 P: E5 }: ]) W! @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：
$ @$ E1 o3 @/ \' W& t* J5 K, B3 f% w[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
. u# F7 [2 \$ X8 |3 q6 qa=postmnmx(an,mint,maxt);
7 ?; L  z5 c& Z/ V2 m: Lpnewn=tramnmx(pnew,minp,maxp);
anew=postmnmx(anewn,mint,maxt)
我修改为：
0 Y! w/ `& S, n) T( q[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
9 e! {6 M) O! W& ya=mapminmax('reverse',an,ps2);
  I- \8 B7 M# ?' n% i3 A! y7 upnewn=mapminmax('apply',pnew,ps1);
anew=mapminmax('reverse',anewn,ps2)
- E8 k7 Y" r& W2 h- Q: X6 l# D
* F5 B5 n3 W/ z; X6 a4 M3 g原程序为：
%原始数据输入
; Q4 q$ ?+ l* [* Pclc
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 q+ D' P* \/ M" h    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,...
+ Q4 N. h! l1 S( l1 q1 O9 R    2.5,2.6,2.7,2.85,2.95,3.10];
8 I# m) N  E/ C# N& H/ V: Qsqglmj=[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];
: Z0 P4 R2 ]+ M' e3 l8 [  ~) z, n5 |glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
2 m; k- ~- R, q6 k6 y  _  z1 d0 w    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

%数据归一化
[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
dx=[-1,1;-1,1;-1,1];
. N! O) d' t6 M+ D; Y! T
%BP网络训练
net=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
net.trainParam.Lr=0.05;
$ s3 _  E/ b! O/ X# M* Ynet.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
, d& l& r8 {7 @5 v5 j
%利用原数据对BP网络仿真
2 e$ ~1 w5 S$ H$ k0 A: l# Zan=sim(net,pn);
; U- Q2 e; Z" J- n$ Q* v5 K% z. d" Na=postmnmx(an,mint,maxt);
+ P+ r, Q/ Z7 e" z3 e' \
$ @7 i( d# s# v8 T  S%仿真结果与原数据对比测试
/ j$ x! P& f% {3 g1 O+ ~x=1990:2009;
newk=a(1,;
newh=a(2,;
figure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
. H0 d$ E5 o: o+ f- Hlegend('网络输出客运量','实际客运量');
xlabel('年份');ylabel('货运量、万人');
( q$ l* b% g2 dtitle('运用工具箱客运量学习和测试对比图');
0 u* W/ {5 w0 G( O* msubplot(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');

# K: b+ B! Y4 |  J%新数据仿真
pnew=[73.39,75.55
    3.9635,4.0975
    0.9880,1.0268];
# Y  f7 ~$ }# }' n4 Tpnewn=tramnmx(pnew,minp,maxp);
anewn=sim(net,pnewn);
anew=postmnmx(anewn,mint,maxt)
0 t6 p: u2 i1 B
修改后程序为：
& J9 }! f3 R& [% K%原始数据输入
6 e  }7 l4 b: x1 }: Qclc
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,...
( Q, q% `$ u& Y    2.5,2.6,2.7,2.85,2.95,3.10];
% g' p- M0 q! `7 e& l! V0 d' xsqglmj=[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,...
" F. O! \. D7 w" ^1 X7 S3 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,...
+ y; `$ E2 @7 Z. u    22598,25107,33442,36836,40548,42927,43462];
glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
, Q' L# A* h6 Y( N8 J/ b* Q    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
- d) L* s* Y+ T8 N0 q* ^) Et=[glkyl;glhyl];
2 r$ M$ ~0 h2 e1 o2 f2 m0 m' k
%数据归一化
. {' Z3 G. v* l% x  l0 M  a' Q8 ]* h[pn,ps1]=mapminmax(p);
% q3 P( z0 v/ J: u[tn,ps2]=mapminmax(t);

%BP网络训练
  G3 u+ l; ~1 O5 @net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
+ F. z  e/ e, G' r% n# M, wnet.trainParam.show=1000;
net.trainParam.Lr=0.05;
2 l  \3 k+ M% ^6 u) [net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
/ a: F6 q% y" E
%利用原数据对BP网络仿真
4 c! [, Z1 n: w. Pan=sim(net,pn);
8 b! o# {6 j+ Ia=mapminmax('reverse',an,ps2);
9 c- B( m3 Y; U' k' U# d
%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,;
newh=a(2,;
figure(2);
7 ]; L4 V1 O5 n- ^6 n, c# Gsubplot(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--+');
0 b, l6 d2 d, `. dlegend('网络输出货运量','实际货运量');
xlabel('年份');ylabel('货运量、万吨');
* a/ k- f0 I5 w2 \6 ?. x! r8 btitle('运用工具箱货运量学习和测试对比图');
2 ^- Z. k% m% P: ^
" ]1 ]4 L$ u; p' b' u$ I%新数据仿真
pnew=[73.39,75.55
    3.9635,4.0975
- z" Y; A1 F7 N, f- a% K$ p& ]    0.9880,1.0268];
4 v, m) \5 q( y; k  Ypnewn=mapminmax('apply',pnew,ps1);
' V2 `- R& f/ {- eanewn=sim(net,pnewn);
; g) _0 y% B7 h) N3 {9 janew=mapminmax('reverse',anewn,ps2)
8 k. [2 q# P+ w& f（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; u8 t% k( @# g\" W; d
   
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,...
   % h4 ?6 \5 g$ c$ j
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   . [& U  u% Z, g4 p) O4 V5 \
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 ?# u2 B9 L- Q& N1 M
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];5 b8 r( l: n( 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,...
   3 N& c! r9 i. H% p
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   * A. G9 d9 i* z
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   9 g  k' h7 J5 j' t! Z) R
9.      22598,25107,33442,36836,40548,42927,43462];
   5 l$ M: r* |, A, {5 q
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    + X; D+ }8 G! U, S
11.      13320,16762,18673,20724,20803,21804];& A8 t\" s; ]+ p: C
    
12. p=[sqrs;sqjdcs;sqglmj];
    4 [\" _) W+ a* f; n: h. _7 T
13. t=[glkyl;glhyl];
    & j. O& C  f& o# o$ d, F; u# U
14. %数据归一化
      b# p! X) m2 h( H& [' b/ C2 W
15. [pn,ps1]=mapminmax(p);6 y! G# J: S6 o  o( I9 h
    
16. [tn,ps2]=mapminmax(t);
    9 H8 T0 k  N( m# V9 d
17. %dx=[-1,1;-1,1;-1,1];& _1 t6 M& t. B/ d7 m4 V% ]
    
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]( Y+ i0 F6 w9 w  T* V9 Y; P) v
    
19. / t0 ^- {6 t' X. ~
    
20. %BP网络训练
    & T: A) `, n9 Q- P\" P
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]
    ( R/ ~+ e- K$ h$ }. ~
22. net.trainParam.show=1000;/ c3 i3 [' t1 u+ S7 C' A3 X
    
23. net.trainParam.Lr=0.05;0 r! {3 \\" `, W1 Y$ J
    
24. net.trainParam.epochs=50000;- \' b. ]9 f2 o% t
    
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]8 }- p; b9 `! L0 @: P
    
26. net=train(net,pn,tn);
    \" _; {$ s, Y\" e
27. 
    / j/ D: N' Q) Y5 F1 d, C
28. %利用原数据对BP网络仿真. g\" w6 M% }/ R/ T
    
29. an=sim(net,pn);
    8 d% e/ Y1 V, |4 g! {3 {
30. a=mapminmax('reverse',an,ps2);0 b+ y5 ?; s: C5 r1 F! P% A6 A
    
31. ( A, Q7 t/ @2 D; @1 n* ~+ H8 P
    
32. %仿真结果与原数据对比测试 - s7 t8 ^& _/ D2 `% J
    
33. x=1990:2009;\" \# q6 B2 U/ I/ \  H
    
34. newk=a(1,:);
    5 W\" f; p2 P4 V( W6 o/ S
35. newh=a(2,:);
    0 ~. o' N0 n0 e* I( J
36. figure(2);* Y7 N2 h0 L0 m- R; e3 O
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    : n' [3 J' M3 n$ r5 G4 I
38. legend('网络输出客运量','实际客运量');
    , z$ k+ R: J: w' b9 i\" k
39. xlabel('年份');ylabel('货运量、万人');2 Q0 p0 Y# Y) [1 S; ]! u% G! u1 j4 y; C
    
40. title('运用工具箱客运量学习和测试对比图');
    9 W! x) B4 |: p% z/ Q0 b) `* D
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    + }. r! R- W5 Z( d
42. legend('网络输出货运量','实际货运量');6 S/ e' z\" W/ O3 N
    
43. xlabel('年份');ylabel('货运量、万吨');6 Q: P; M0 g/ F! m' W
    
44. title('运用工具箱货运量学习和测试对比图');9 f& H6 a; Y9 ^( `) A$ y
    
45. 1 K8 ]- `8 z, F! }
    
46. %新数据仿真
    ; [4 o8 U\" V! x
47. pnew=[73.39,75.554 i# r# |7 Z4 M% t+ ~
    
48.      3.9635,4.0975' ?/ x7 K/ I0 a' s0 X
    
49.      0.9880,1.0268];2 \) ]+ w8 e5 X  K. @
    
50. pnewn=mapminmax('apply',pnew,ps1);
    8 W$ _# g3 b3 h) X, L
51. anewn=sim(net,pnewn);5 ?% @0 p; l3 L
    
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$ A3 s$ R; |, [1 R- g
   
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,...
   7 @7 E2 y2 K\" Z9 u0 X0 `% R
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];& r' U9 e. d) X6 Q* P) F
   
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 B# ^7 C+ x7 G
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   \" W4 \. m! K8 Y: x2 O/ \
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,...
   2 p) O  ^5 s, @- r
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   & i% e) i9 r' O9 ?; X7 g
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   * Z9 k% W5 i  @' v4 O
9.      22598,25107,33442,36836,40548,42927,43462];
   5 E) O( D$ G3 b2 g- V/ I  _
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    5 h: O4 j5 h* P6 ~$ n0 `4 W
11.      13320,16762,18673,20724,20803,21804];
    # T5 r5 u2 o+ V7 R' Z6 \8 e
12. p=[sqrs;sqjdcs;sqglmj];
    3 z: J7 D8 P6 H) O\" o0 `
13. t=[glkyl;glhyl];
    1 M9 \- H6 }5 w& m) Z7 U
14. %数据归一化
    + [' Z5 D9 _3 s, k
15. [pn,ps1]=mapminmax(p);
    1 g) X9 M$ ^  _
16. [tn,ps2]=mapminmax(t);  O! i' U! F! q# ~* m
    
17. %dx=[-1,1;-1,1;-1,1];
    * `7 ?' I7 v\" e8 H7 M  x
18. p0=minmax(pn);t0=minmax(tn);
    ( `8 W! n( n1 l
19. 
    : Q8 G: L9 h( C  I  |
20. %BP网络训练
    7 F& K$ X) M% w- p9 W/ I
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
    0 A2 y/ [2 ?7 w$ Q
22. net.trainParam.show=1000;
    + Z% Z9 ?0 [5 v1 |8 O0 b7 ~
23. net.trainParam.Lr=0.05;# O4 c* o4 O9 F4 |$ z- {
    
24. net.trainParam.epochs=50000;\" j$ R2 A7 q# C7 _* a$ x, e
    
25. net.trainParam.goal=0.65*10^(-5);( w5 W+ G4 o\" o! ~
    
26. net=train(net,pn,tn);
    , i1 y3 d! |/ W
27. 
    , X: F- [' t9 h: [
28. %利用原数据对BP网络仿真
    + ?$ ~/ E# Z) n
29. an=sim(net,pn);
    8 |3 [7 e) l9 _# A6 q. m
30. a=mapminmax('reverse',an,ps2);( [( d, {; {8 ~2 y$ a: h) g
    
31. 
    , x# }5 T1 i5 z  R% k
32. %仿真结果与原数据对比测试 8 B: o2 J5 H9 R! a* ~\" t
    
33. x=1990:2009;! D/ [+ G3 L, {7 _
    
34. newk=a(1,:);, h# s2 ?4 R' ^1 Y) x& q$ }
    
35. newh=a(2,:);* l6 ^5 f5 B% N% ]
    
36. figure(2);  o% y6 z! O$ l& n( G\" G
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    . L* B, @0 }/ r
38. legend('网络输出客运量','实际客运量');, `. w1 h; z! G2 v
    
39. xlabel('年份');ylabel('货运量、万人');
    ' U0 F9 g/ I5 s0 |* g: |0 l( z
40. title('运用工具箱客运量学习和测试对比图');, _% \3 D* X\" g. e
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');$ V' o* W. i, K: ?; c* R3 g8 Y
    
42. legend('网络输出货运量','实际货运量');
    \" b5 z! m$ J0 Q
43. xlabel('年份');ylabel('货运量、万吨');, J; A% J8 b3 L+ l6 T
    
44. title('运用工具箱货运量学习和测试对比图');
    $ v, J8 u5 Y  L! I/ b1 Y
45. 
    . _& _9 r- B! w6 a  Y% \$ g7 I% q
46. %新数据仿真: Z' ^4 G/ E: Y9 w' {
    
47. pnew=[73.39,75.55/ \8 M8 h% T) p8 x6 p
    
48.      3.9635,4.09759 e, `2 D0 U* w& C! q1 b) n
    
49.      0.9880,1.0268];; ]$ q$ g6 \\" o/ z
    
50. pnewn=mapminmax('apply',pnew,ps1);) M% m: D6 q8 c; K4 Z  L% D7 ?
    
51. anewn=sim(net,pnewn);# {/ b9 b3 r+ p) ]( `
    
52. anew=mapminmax('reverse',anewn,ps2)

复制代码

T-Eric

且生 发表于 2014-7-21 18:22
3 V9 W+ X& ^( a忽略我上面那个，改动的地方在18，21，25行……求批

/ ~6 s; W9 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
   0 g, }0 h  B5 }# S
2. net=newff(p0,t0,6,{'tansig'},'traingd');
   # r+ s( A7 {1 b# Z, Q3 h\" T$ E
3. net.trainParam.show=1000;% w7 A# l, i/ R
   
4. net.trainParam.Lr=0.05;3 \8 J6 D0 V2 c) i
   
5. net.trainParam.epochs=2000;
   5 P  f  p( Z  Q, ?% v
6. net.trainParam.goal=0.65*10^(-4);5 r$ U7 m; I9 k0 O
   
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
三、迭代次数改成2000
4 x/ e& g8 L0 d) ?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