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！！
" r$ z+ x1 z4 r/ `
! D5 u" ?' M& N) G6 {+ o, e: ^%原始数据输入
clc
8 I) N% j$ r7 d# c/ Ssqrs=[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 n: ^$ ?+ J7 asqglmj=[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,...
- y8 X9 ~1 ]( E8 l- [    0.56,0.59,0.59,0.67,0.69,0.79];
5 }, W( r- y+ z! F$ [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 c8 x# `! \% g6 O( W1 J3 [    13320,16762,18673,20724,20803,21804];
, h) ]9 Z+ g/ }1 ]; q- h6 W$ z/ Lp=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

+ d( h4 w* W9 R%数据归一化
[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
* [* r0 }2 o; {* n8 m- O1 ydx=[-1,1;-1,1;-1,1];

%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
1 u' Q' Z, e/ F4 p; nnet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
3 U  P1 G) t) x' ~$ w
1 t$ v: D5 t# ^3 u%利用原数据对BP网络仿真
3 e# u3 N8 X2 t  Uan=sim(net,pn);
4 P7 }. H$ p2 g  F! f5 b9 ia=mapminmax('reverse',an,ps2);
/ K- u# m  d2 m0 E! k4 C' b; K1 w
%仿真结果与原数据对比测试
6 A2 D6 {; e6 u) p5 ]x=1990:2009;
newk=a(1,:);
newh=a(2,:);
; w* x* s; _0 L; [! j, 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');
, e% Z' b, x1 V. l8 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');
3 |6 O' [: x+ m: }8 _9 a; ftitle('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
    0.9880,1.0268];
+ d: s" q  x# o) K! S" ypnewn=mapminmax('apply',pnew,ps1);
% X) [# n$ ~/ Eanewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
( t0 P9 W" {9 P! A

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
! ^+ d% S% `: w, V9 L/ ~  h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 ...

* o$ v2 j& S1 g, R$ I% i& N. [因为版本更新，Matlab中的Newff命令用法有所改变，原命令为：
/ ^2 c/ _/ _) X. \# s& _net=newff(dx,[3,7,2]{'tansig','tansig','purelin'},'traingdx');
% X" _/ V2 X" D' [, ^" L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：
' u3 m" t, E* r4 _/ b+ ?[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
a=postmnmx(an,mint,maxt);
/ v! b3 G; T5 o2 I7 D: D+ ]pnewn=tramnmx(pnew,minp,maxp);
anew=postmnmx(anewn,mint,maxt)
我修改为：
[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
# l/ a, N* w- O0 ]8 o& T! x4 xa=mapminmax('reverse',an,ps2);
9 {0 |* J! z" P! u3 ]( spnewn=mapminmax('apply',pnew,ps1);
$ X8 q' T9 j5 J5 E4 M4 ranew=mapminmax('reverse',anewn,ps2)

原程序为：
%原始数据输入
clc
7 K$ o4 |1 y! B. h$ ?, v; ^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 w0 [" S5 Z# E- ~& Y- u    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,...
# M( F- ^7 W' ?& `    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];
glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
9 g/ Q2 b* h: p( S4 x" f    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];
, m; h* A) w8 Wt=[glkyl;glhyl];

%数据归一化
[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
dx=[-1,1;-1,1;-1,1];

%BP网络训练
+ @1 D0 m( e! L! e, jnet=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
+ b- w7 E& S3 d+ inet.trainParam.Lr=0.05;
) u& j2 ^8 e- ?+ 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
& v: _, B9 T. X4 Van=sim(net,pn);
a=postmnmx(an,mint,maxt);

& o1 K# I( `( K  V' B$ J%仿真结果与原数据对比测试
" X3 \- `* o# }# bx=1990:2009;
newk=a(1,;
newh=a(2,;
$ {/ P, p- [1 d6 W" n# bfigure(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');
; C+ ]1 Q/ G! R  hsubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
. V  A1 ?6 b3 o0 blegend('网络输出货运量','实际货运量');
9 q2 s* I) W" rxlabel('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 ]" u" |; ]" j
1 @) _% M" U$ r; D* J%新数据仿真
9 y$ r! U% P% e% cpnew=[73.39,75.55
    3.9635,4.0975
* q7 }# f# c" ?& k5 s4 I. W    0.9880,1.0268];
pnewn=tramnmx(pnew,minp,maxp);
anewn=sim(net,pnewn);
4 f: R; ~, d, p5 p& ?5 r% g9 S9 U$ ianew=postmnmx(anewn,mint,maxt)
6 J: f7 u7 n7 s0 J/ O4 c
$ k' _! _: h) j# {  u9 P: u修改后程序为：
%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];
! N- Z' R3 h4 y6 {" a: Q) csqjdcs=[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,...
    0.56,0.59,0.59,0.67,0.69,0.79];
, R% J7 X* D. ~- V0 zglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
) C- K% \8 ~, X$ Y: _    22598,25107,33442,36836,40548,42927,43462];
1 `5 K3 K0 T3 d. j  ~6 Q, C3 Pglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
1 g5 T% K5 H" j! l! b    13320,16762,18673,20724,20803,21804];
7 K/ Q6 _# G- H: k- [5 Ip=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];
8 X$ C) E' u; T$ F4 U  b( z
%数据归一化
. N9 S/ f5 L* _4 e/ m+ C/ s2 F[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
! A5 Y+ C% h5 V0 V* a0 u& H" c
( ?/ [8 |* ~1 v* x) `; K; ?- H%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
9 _2 {9 |; T9 {. P9 Cnet.trainParam.show=1000;
  l% q9 i  R: R# P! N' c8 xnet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
' \7 f9 O! Z$ z! l* o4 R3 dnet.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);

%利用原数据对BP网络仿真
# z# i2 w! A5 t/ C" V' S4 uan=sim(net,pn);
a=mapminmax('reverse',an,ps2);

0 Q7 x' X8 e/ U4 l%仿真结果与原数据对比测试
x=1990:2009;
  `: ^) ~: F2 i1 C! }7 Qnewk=a(1,;
0 V- s# I" K  ~! onewh=a(2,;
- w% e) f2 k1 E6 i5 f& Ofigure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
9 i& N4 ^1 Z: t; X- ?5 `legend('网络输出客运量','实际客运量');
xlabel('年份');ylabel('货运量、万人');
! o; Z% z6 _) c. a5 X; T8 n3 etitle('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 Y* ]7 F1 [4 ^2 M9 xlegend('网络输出货运量','实际货运量');
0 c: c& ^& I( c' 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');
5 @7 w" W" r6 }5 [1 X" d9 P( k
* O4 \9 S" e2 a3 S2 [  c  s, K%新数据仿真
) P& L8 A* ^" `! p7 a! ~0 cpnew=[73.39,75.55
8 h5 S  X+ F& q    3.9635,4.0975
    0.9880,1.0268];
8 d8 e, S( a0 [2 W1 p! H, zpnewn=mapminmax('apply',pnew,ps1);
1 d0 f. j8 @- |. H* |anewn=sim(net,pnewn);
8 p0 J  e4 V- Y+ u5 A5 Danew=mapminmax('reverse',anewn,ps2)
（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
   % j; I; X& s8 b# V  h
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 v' Y8 a  ^8 F1 L: x) Z4 S; Q
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   * r( G1 }: M% `
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 O/ Y/ T. ^0 R/ ?( i4 F6 c% n: A
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];* i. j% ~) F7 N/ N$ W% @! 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,...+ H# h: y. B  k) i. J# ~
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];9 f1 l1 Q' U' p6 ]
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...9 S$ d8 H3 R4 b2 R\" `
   
9.      22598,25107,33442,36836,40548,42927,43462];1 C! H$ x! K) n9 r; }
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    : K\" |1 D! x\" V$ u/ ~. e, v) |
11.      13320,16762,18673,20724,20803,21804];
    1 T. A5 U9 [0 G( u& r& o
12. p=[sqrs;sqjdcs;sqglmj];% Y\" |\" E) y0 f
    
13. t=[glkyl;glhyl];) C& X* ~. [8 ^3 v
    
14. %数据归一化
    1 H0 g2 \; X# M\" {  v7 {
15. [pn,ps1]=mapminmax(p);
    + P  d; N5 Y% U7 d( D. r
16. [tn,ps2]=mapminmax(t);
    8 {4 B4 \+ A, h
17. %dx=[-1,1;-1,1;-1,1];- D* D- g# t2 m\" J4 A! {, C: _
    
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]
    7 v% k% T/ p) H5 @- ~( S/ _
19. 0 U) X6 V7 t# ]( n
    
20. %BP网络训练
    & t+ Y6 [$ B; E# V1 c- S
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]
    5 d8 h* @- X- k
22. net.trainParam.show=1000;) N. x, D' [9 O$ c5 N# l' Z
    
23. net.trainParam.Lr=0.05;: J- g0 z% y9 R
    
24. net.trainParam.epochs=50000;% J( b8 z! ~& G' j
    
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]
    ) R) b$ W/ L0 ~
26. net=train(net,pn,tn);
    7 T) ]+ v# s! _
27. 
    1 A4 _* s1 o5 t
28. %利用原数据对BP网络仿真
    . P/ r: g9 C8 R7 ?
29. an=sim(net,pn);8 R0 S# T' u& `5 I8 g3 T
    
30. a=mapminmax('reverse',an,ps2);
    - _4 }& A% g\" U7 W3 V
31. 
    8 B8 O8 B& ?# U. ~6 w* ]* P; P
32. %仿真结果与原数据对比测试 2 ^6 t) Q5 w9 c6 S/ X# o7 K
    
33. x=1990:2009;2 G. W+ U- `) ]
    
34. newk=a(1,:);7 V) Q; h: R9 _5 Q
    
35. newh=a(2,:);
    ' [0 C7 C1 m- E4 I5 B5 i
36. figure(2);
    ! r1 ~$ `) G\" u9 i& k
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    : v! @  g2 o7 D9 P
38. legend('网络输出客运量','实际客运量');' ]2 L4 h: L3 w, F; K6 F9 L
    
39. xlabel('年份');ylabel('货运量、万人');
    2 I, `2 ^# `; e! N' ?1 y
40. title('运用工具箱客运量学习和测试对比图');* x1 R$ s/ x5 E9 z9 Q
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');2 M) T! M. A* Z$ O1 Q/ T
    
42. legend('网络输出货运量','实际货运量');
    \" X+ A. d; {7 c# ^2 L, T) i
43. xlabel('年份');ylabel('货运量、万吨');
    & I9 }$ q( v# v$ c# z' A: R
44. title('运用工具箱货运量学习和测试对比图');' l. q; v6 E! z
    
45. 8 m8 ]  ~' U/ ~7 P) z- }4 N+ t) H
    
46. %新数据仿真, D0 {( k; E/ A( Q
    
47. pnew=[73.39,75.555 ^\" I5 [$ A) d0 ]& x/ s, [9 }0 [
    
48.      3.9635,4.0975$ z/ k1 A% E# x3 w+ E: g. i7 O
    
49.      0.9880,1.0268];% q6 @\" U2 u; Q6 f
    
50. pnewn=mapminmax('apply',pnew,ps1);5 N8 q$ T' f, I9 x. T& g
    
51. anewn=sim(net,pnewn);
    - I2 K+ n! k\" q9 V
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. clc8 K! v9 C6 J. p& u  G. B
   
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,...
     u) N* b0 b7 \
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];1 I9 S$ S8 J& H. s4 u# J\" \
   
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,...
   6 s$ U\" M! J1 k, e, W  Y6 f
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   8 l# J  y9 Y& O/ L\" 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,...
   & z  ~5 u5 T; Y5 j9 w9 L% W
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   ! a; U8 S) s' D
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...6 ?* C' H9 X, E) a8 v% G+ ]. P5 \
   
9.      22598,25107,33442,36836,40548,42927,43462];  i% y! G+ @, o* B% l
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...- j$ y; O7 \9 C% h# y# V
    
11.      13320,16762,18673,20724,20803,21804];
    7 K6 C- d0 p8 Z6 A8 _9 y
12. p=[sqrs;sqjdcs;sqglmj];* n( N) |+ N8 f9 s) b. P$ Y
    
13. t=[glkyl;glhyl];8 |+ r8 L; L5 l4 I, T+ p
    
14. %数据归一化8 T# U+ t+ p4 z: B7 s, D2 U0 \( i
    
15. [pn,ps1]=mapminmax(p);
    / q3 {7 a9 Q- p9 W
16. [tn,ps2]=mapminmax(t);
    9 w) I3 f' [# Y. F/ z
17. %dx=[-1,1;-1,1;-1,1];) R) X\" _\" c5 ]5 ?
    
18. p0=minmax(pn);t0=minmax(tn);; |, _) m; p# C
    
19. 
    ! y$ F) R. V& }* f' p9 h
20. %BP网络训练# ^/ F( C  C4 B! {# O$ T8 w5 Y4 Q
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); 0 ?9 _) K. \1 v& Y/ b
    
22. net.trainParam.show=1000;& L1 c. e- K  p& X$ P, N. P9 f+ i
    
23. net.trainParam.Lr=0.05;
    9 r; _3 Z1 \* G! i- z  u& e
24. net.trainParam.epochs=50000;
    ; C+ n6 f/ u3 m# G8 T; m
25. net.trainParam.goal=0.65*10^(-5);
    / Q: {& w$ q) C5 F! K( f
26. net=train(net,pn,tn);9 d7 e+ U- u6 z\" F; L# w
    
27. , H# H0 E/ l: V) G1 e
    
28. %利用原数据对BP网络仿真6 L% o& v/ t! q+ `
    
29. an=sim(net,pn);
    : V% _! M2 j+ X
30. a=mapminmax('reverse',an,ps2);' }3 o4 E\" {& t4 o* p% ~' v
    
31. 0 r( O, ~$ v; N) X/ V
    
32. %仿真结果与原数据对比测试 * B* @4 V* R# F! M- W6 P0 m/ U
    
33. x=1990:2009;
    : Q# I. e( G* l/ Y
34. newk=a(1,:);9 h8 M: c8 N9 W- ~6 N
    
35. newh=a(2,:);8 N  N9 Y& e( n: C* q# z+ ~
    
36. figure(2);
      y' I1 X, p$ e5 h  c2 \! |0 z\" q
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');( \$ J\" U! t$ Y; b
    
38. legend('网络输出客运量','实际客运量');8 z5 o4 W, R5 H4 \6 {2 q  |
    
39. xlabel('年份');ylabel('货运量、万人');% T6 `9 z) P7 L\" i; t
    
40. title('运用工具箱客运量学习和测试对比图');' Z& _0 l1 q6 I# w\" m% e' t\" ^  n
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    ( F' T4 A' d4 P, Z% u& c
42. legend('网络输出货运量','实际货运量');8 V( G9 Y, y; X0 n8 h8 \, ?
    
43. xlabel('年份');ylabel('货运量、万吨');- ^% k/ S# \: |1 @5 D& s5 Z) }
    
44. title('运用工具箱货运量学习和测试对比图');
    9 |0 \* V; }4 Y; v6 o( @( D  N
45. 5 W5 U6 X/ c3 W6 d
    
46. %新数据仿真
    . }3 B+ H! @\" U0 k
47. pnew=[73.39,75.551 ?( I2 Y. F$ e/ F: b- m, m  O
    
48.      3.9635,4.09752 S' ]* c3 Q+ Q' G4 n4 _! o
    
49.      0.9880,1.0268];
    1 n, R3 Q- Y: Q8 e6 I+ P) V& R
50. pnewn=mapminmax('apply',pnew,ps1);
    \" O2 C4 r9 H\" _
51. anewn=sim(net,pnewn);
    6 Q- s  R$ K! X) Z
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

' {6 w, k  z3 t' a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
   \" _& v9 Q' `! ?3 K5 d6 {
2. net=newff(p0,t0,6,{'tansig'},'traingd');
   7 c4 [1 t. ^/ U
3. net.trainParam.show=1000;* Q8 w- i7 h0 k# Q& U\" X
   
4. net.trainParam.Lr=0.05;9 n) [$ A( w2 C
   
5. net.trainParam.epochs=2000;3 @( E( T; _\" z0 Y8 J
   
6. net.trainParam.goal=0.65*10^(-4);
   3 m- W* l0 N& Z% Q) p
7. net=train(net,pn,tn);

复制代码

我只改了这里面的，
+ n9 {) H7 O1 X& t4 S8 y" `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
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