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！！

%原始数据输入
$ N+ M% Z1 `4 m+ q( S* s6 _! lclc
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,...
. F' M8 w9 ?9 V, `) C9 k    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,...
7 f3 o" q% U6 ~3 r9 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,...
# i; E( S1 s/ F% L    22598,25107,33442,36836,40548,42927,43462];
glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
& C- x3 n: e: E/ {+ Y% h# j) }3 Y' N    13320,16762,18673,20724,20803,21804];
5 K6 [# U! u, S9 [/ ap=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

. s9 y$ Y. n1 q1 ~%数据归一化
[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
dx=[-1,1;-1,1;-1,1];

%BP网络训练
; Y5 D) t1 r8 K" ~: C+ J% t3 |net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
/ R) }% n5 b6 G! ^& ~net.trainParam.show=1000;
net.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
2 ~7 G8 }6 k9 Unet.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
2 ~# D; R" U! [* }5 U
%利用原数据对BP网络仿真
an=sim(net,pn);
: |- J# J& t5 sa=mapminmax('reverse',an,ps2);
- b1 E3 f8 y0 _. |9 A6 t+ d( L) Y
/ N( j, d+ n. o, c2 u%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,:);
9 p4 ~: s" t: ~& Znewh=a(2,:);
figure(2);
  T" q; n4 @$ t  I9 C, L, E6 C% E8 [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');
/ n9 @1 t+ k" n8 @xlabel('年份');ylabel('货运量、万人');
# W% O# y6 M! s) T/ ^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');
7 b8 e) J- S! U" N2 y! ~& d% Mxlabel('年份');ylabel('货运量、万吨');
3 ^; ^1 u# Q& ^title('运用工具箱货运量学习和测试对比图');
- j& k. f. f3 j; W
%新数据仿真
pnew=[73.39,75.55
    3.9635,4.0975
    0.9880,1.0268];
' n1 X1 x) I( Vpnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
0 e2 E6 d8 Q: d/ k

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
( Z" y8 }, q) D% U0 W! Z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 c# f! ~6 P" w& I4 B( i9 f因为版本更新，Matlab中的Newff命令用法有所改变，原命令为：
) f( J" O$ k/ v) y; w1 o7 inet=newff(dx,[3,7,2]{'tansig','tansig','purelin'},'traingdx');
我修改为：
! k4 i/ n# a% I6 Snet=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：
[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
a=postmnmx(an,mint,maxt);
pnewn=tramnmx(pnew,minp,maxp);
anew=postmnmx(anewn,mint,maxt)
我修改为：
[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
" Y' h5 z( I5 w8 k$ }% ?: R: ia=mapminmax('reverse',an,ps2);
& Y; o1 [$ y8 I* K+ z9 i+ Zpnewn=mapminmax('apply',pnew,ps1);
anew=mapminmax('reverse',anewn,ps2)

* e" P0 y/ I, g: K& t; \7 Z( R$ B8 o原程序为：
$ S5 R* y; T! n$ g% h) o%原始数据输入
( G8 p; ~, Z# h. lclc
% M& D" \7 E$ Q" Dsqrs=[20.55,22.44,25.73,27.13,29.45,30.1,30.96,34.06,36.42,38.09,39.13,39.99,...
" w6 s) }: C: l9 w, R$ i9 y    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,...
, r9 f. [) d% p, d: u: W0 v    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];
9 `5 i8 m6 C" }0 r6 Yglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
" f) F0 q* R* d( w3 w# ]    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];
. A0 `: e7 E# {) L- [/ Rt=[glkyl;glhyl];

%数据归一化
& N, [3 ~) `7 f. u: E5 [6 K[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
0 p- e; o( m) [0 \dx=[-1,1;-1,1;-1,1];

/ b" K3 W: t. e9 I5 w! J1 h; Q- T, u%BP网络训练
net=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
net.trainParam.Lr=0.05;
# d7 z% y" V5 {0 {3 A, f5 Anet.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
an=sim(net,pn);
a=postmnmx(an,mint,maxt);
" b% ?6 {+ r* ?1 _
%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,;
$ A& E5 y. m" G: Bnewh=a(2,;
0 ~  W3 c- F; Qfigure(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');
2 u' t9 Y. r# G; Y$ fxlabel('年份');ylabel('货运量、万人');
+ q; L: l3 n: E1 {  ^title('运用工具箱客运量学习和测试对比图');
/ p2 U! A! I0 }' f! W; z1 p. n/ ?subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
( w: \+ C& G( W5 B* ^/ qlegend('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');
' D1 S! F+ ^; b$ u, o3 L/ x4 _7 o' d
1 \* h! v5 h4 E0 d3 M# z! I. j%新数据仿真
pnew=[73.39,75.55
    3.9635,4.0975
7 Y+ {& ~: q+ X# ~% d. q& M9 \8 Q    0.9880,1.0268];
pnewn=tramnmx(pnew,minp,maxp);
anewn=sim(net,pnewn);
anew=postmnmx(anewn,mint,maxt)
$ U) r% c" g: x/ Y& h% h: G
修改后程序为：
$ Y8 i- t- H$ O%原始数据输入
clc
' i3 }4 z# G/ V5 `7 gsqrs=[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];
9 R1 h, F* `3 u, d: [7 fsqglmj=[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 ^8 q& f  L& A0 _- b* E+ v0 x2 w1 O    0.56,0.59,0.59,0.67,0.69,0.79];
; f& [- X# X- a1 R* ^0 iglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
0 _- X- s$ ~7 J7 w6 C4 F8 M    22598,25107,33442,36836,40548,42927,43462];
" J( I5 W/ m; h1 @! Pglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
1 Q) ]4 m$ H# f+ W    13320,16762,18673,20724,20803,21804];
0 H; F: Z7 r7 A4 ?$ P+ K9 }! }p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

  G* @- q$ B6 c6 R# g! M! E1 r3 C%数据归一化
: l+ @7 g9 n$ k2 U  R- o- q/ }[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);

0 W# a- f. g" L# ^  z%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
4 j* ?# G( O7 C7 l: J5 m5 m- H' cnet.trainParam.Lr=0.05;
6 U5 i: t# l3 q) C. [net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
/ j. D1 _+ T! f& A7 D! C+ C7 u3 W
%利用原数据对BP网络仿真
# C4 m, @: }! l" dan=sim(net,pn);
a=mapminmax('reverse',an,ps2);

5 O; \" F5 u. _; W4 o1 S1 X! n%仿真结果与原数据对比测试
6 p2 s$ Z- y. B( y2 d5 P+ Jx=1990:2009;
newk=a(1,;
) h( L* y* j9 I  L  q( N0 Pnewh=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');
8 }+ u1 E/ T8 pxlabel('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');
" Q8 V; w) l1 b; b! dsubplot(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');
  z# L0 l- M* _7 ]% ]/ Ttitle('运用工具箱货运量学习和测试对比图');
% }+ f+ ^9 E4 b2 F
%新数据仿真
" ^& H$ }) z  N8 i6 C9 fpnew=[73.39,75.55
& o5 H" x) R0 @    3.9635,4.0975
    0.9880,1.0268];
8 k0 U' \+ Z& v* h9 {8 ~pnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
1 w4 j: V8 w8 r% {anew=mapminmax('reverse',anewn,ps2)
（修改的地方用颜色标记了）
' c9 n- o6 F& l% c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. clc1 h2 r& n\" B  |  t! u8 l6 m  {
   
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,...
   0 d5 f, ]3 U3 U0 r
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];\" L6 t  w3 ]3 u, Q5 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,...
   . {) F7 q& E& G0 ]+ q
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   0 @3 c& X. J0 E; B\" |4 W
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,...& w5 u( I6 h1 ^
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   + h& p0 F: [2 L4 O! r8 a* T
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
     S2 U/ K1 D& o. q\" C
9.      22598,25107,33442,36836,40548,42927,43462];7 X! g# k$ x; w
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...4 }0 K% }/ T: _/ i2 L, u0 x2 d  w
    
11.      13320,16762,18673,20724,20803,21804];
    ' Q8 M  S4 F6 T+ D- _5 m$ K
12. p=[sqrs;sqjdcs;sqglmj];0 R3 l7 h' b1 K3 x
    
13. t=[glkyl;glhyl];
      f( I, o( G\" {7 n# L% [4 L
14. %数据归一化: N) v- c  H- G$ v
    
15. [pn,ps1]=mapminmax(p);) H0 F, }% s  C  o5 N% O
    
16. [tn,ps2]=mapminmax(t);
    ; R& Y+ M+ ^$ w) q3 I  F
17. %dx=[-1,1;-1,1;-1,1];8 d9 m' c2 U6 D7 d: e
    
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]
    2 C: h! b! l  p\" d\" v4 m
19. 
    - u) Y; S. R& {\" Q
20. %BP网络训练
    , u7 d$ C6 c0 p
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]
    ) s$ f& P4 m  w8 B
22. net.trainParam.show=1000;/ @# V* s( {* C5 i$ E9 L- t
    
23. net.trainParam.Lr=0.05;* Y% {  ~3 X1 B
    
24. net.trainParam.epochs=50000;
    0 {# C% G& s  z! u3 n2 j
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]5 w- k\" Z7 T' C7 ^
    
26. net=train(net,pn,tn);9 h3 T  Y' s5 i; Y2 a
    
27. 0 A* S0 m/ D$ y; o9 j
    
28. %利用原数据对BP网络仿真& m1 ?( o, L- l
    
29. an=sim(net,pn);- `6 E\" p\" w% h7 ~\" P
    
30. a=mapminmax('reverse',an,ps2);( Q8 L7 O, I! p
    
31. 
    / s  j6 z8 x8 f# U  P) `
32. %仿真结果与原数据对比测试
    / ^: e- ^& O\" x4 M( I
33. x=1990:2009;
    1 x4 P' S! K' w( Z$ l
34. newk=a(1,:);& v& a3 g$ G' ]) M# X+ o
    
35. newh=a(2,:);4 v5 S2 E7 F) A3 p) y- k, T6 O
    
36. figure(2);
    7 b. ?+ ?, y\" T: ~
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');, Z9 U9 S  @1 N  J, R3 [
    
38. legend('网络输出客运量','实际客运量');5 t! c0 u# G% m7 s4 W\" ~* W
    
39. xlabel('年份');ylabel('货运量、万人');
    & t5 q/ U0 ?& t8 T0 V
40. title('运用工具箱客运量学习和测试对比图');
    ( S: m7 `& M, `& X; s6 G
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');+ Q6 B' p7 e, l4 K' }\" A
    
42. legend('网络输出货运量','实际货运量');# v6 L. _) m) [, E* f2 {\" E5 T; g
    
43. xlabel('年份');ylabel('货运量、万吨');
    & ]9 ~, J+ E$ T- c5 |
44. title('运用工具箱货运量学习和测试对比图');% ~0 W! ]4 @) [1 v6 |% ?1 Y
    
45. 
    6 h8 D# q1 q4 k9 G\" l
46. %新数据仿真
    . u+ ]$ M% k& ~
47. pnew=[73.39,75.559 O: a! U8 Y8 y1 }! L9 R  d* E
    
48.      3.9635,4.0975  ~. A* J. H5 g2 U: t) z
    
49.      0.9880,1.0268];4 d/ J8 ]. R! a1 @$ M6 b\" B
    
50. pnewn=mapminmax('apply',pnew,ps1);5 c% @2 C3 F3 v$ h9 W( n
    
51. anewn=sim(net,pnewn);' M4 E! h( q\" y) B
    
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
   3 x( w\" X) h5 r
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,...
   ' M; C3 B7 l0 ?( `+ j; J, m. O\" h2 w3 p
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   . V+ g+ ^8 N/ z$ 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,...
   % _- k. o\" B# h
5.     2.5,2.6,2.7,2.85,2.95,3.10];0 b0 d4 `, `0 q, _# @1 D/ t0 F6 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,...
   # h# q% v7 u& K$ V
7.     0.56,0.59,0.59,0.67,0.69,0.79];3 k! Z9 h' z9 {; o6 V; L1 C3 `
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   3 _8 `9 C\" J, r, Z  I\" C0 [) H2 W
9.      22598,25107,33442,36836,40548,42927,43462];
   , V0 P\" ]+ @\" }5 l
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    5 i5 \( H4 L7 p
11.      13320,16762,18673,20724,20803,21804];3 |+ ?; O5 O$ W5 q6 B* _
    
12. p=[sqrs;sqjdcs;sqglmj];2 ?; A# X( f' o3 `, \
    
13. t=[glkyl;glhyl];
    # I% s9 C3 \* s
14. %数据归一化
    ) J0 v! r) ^' q
15. [pn,ps1]=mapminmax(p);( p4 T: [$ C% c\" C& j\" u) c
    
16. [tn,ps2]=mapminmax(t);
    1 ?8 B1 M5 b. M2 O) W2 }( _3 m' J
17. %dx=[-1,1;-1,1;-1,1];4 p$ \: [$ Z( Q, L4 r6 x3 F
    
18. p0=minmax(pn);t0=minmax(tn);
    7 y1 [; }. p% N, e. o5 q
19. 
    3 x# U, ~, x3 q2 `+ ]+ y
20. %BP网络训练$ [6 \! t# x  h- w1 S( {
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); 1 v+ ~8 V+ ?\" w0 s
    
22. net.trainParam.show=1000;
    9 J0 y, q2 {$ X& y; u3 k% P
23. net.trainParam.Lr=0.05;, ~. P' H2 V: r2 Z( D! x# l& m( g
    
24. net.trainParam.epochs=50000;0 q, {+ e2 y2 Z# z# a1 L' F* D7 E- b
    
25. net.trainParam.goal=0.65*10^(-5);+ w1 e, Q' v- J( v. s
    
26. net=train(net,pn,tn);9 n9 s- x# g- B; k0 {
    
27. 
    4 I7 T* V% ~6 t$ N( k
28. %利用原数据对BP网络仿真
    8 x5 u5 ~% B: Y( |9 K
29. an=sim(net,pn);1 U7 i4 S. B\" s* G! G9 B
    
30. a=mapminmax('reverse',an,ps2);: _\" J2 N1 c. d1 G
    
31. ) y% \  A\" A) y% l
    
32. %仿真结果与原数据对比测试 & g. h+ a; Q. Z+ G) B1 l0 L. n, [+ L
    
33. x=1990:2009;3 f' |) i- A7 t  h
    
34. newk=a(1,:);$ K- e1 L6 {# F; o
    
35. newh=a(2,:);
    % U8 G  H5 x6 n: e0 W. }( n
36. figure(2);\" z9 b6 g4 j4 e% U' t& g) v0 g
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');6 `* h/ I5 D- o: N) W
    
38. legend('网络输出客运量','实际客运量');
    - S! J, S2 ~% `5 }4 |) u* D3 t
39. xlabel('年份');ylabel('货运量、万人');
    , r+ X7 C4 P* b
40. title('运用工具箱客运量学习和测试对比图');
    1 i8 n\" q% y5 c# V\" J3 ^* |4 b$ ^
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');: J( ?' P, d/ W
    
42. legend('网络输出货运量','实际货运量');
    + e, |) e' G- @: r
43. xlabel('年份');ylabel('货运量、万吨');; K# [: }( k$ |2 n- Z8 k! Y! R
    
44. title('运用工具箱货运量学习和测试对比图');
    7 p5 Q9 f8 ~9 u
45. 
      B\" {4 K0 ^) Z: J# G% p
46. %新数据仿真
    ; [+ ~, |2 a+ A% m/ M8 W
47. pnew=[73.39,75.55
    \" b1 f  p; |: D4 \
48.      3.9635,4.0975
    * R: F8 j' T. p/ F; t3 I6 }
49.      0.9880,1.0268];
    7 A  ~! K9 c* @0 ]) g
50. pnewn=mapminmax('apply',pnew,ps1);
    3 e$ o- Z+ ?- L3 L
51. anewn=sim(net,pnewn);; X+ B; s\" k( W  M
    
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

* j2 N7 H1 l# K' Q+ }* B' ?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; f! G1 O- j4 \9 V$ R& M
   
2. net=newff(p0,t0,6,{'tansig'},'traingd');
   ; S6 p\" {) G8 M& N8 U- }
3. net.trainParam.show=1000;' z1 X) e2 S. M5 i
   
4. net.trainParam.Lr=0.05;
   / w' C* {: f( `. q2 r
5. net.trainParam.epochs=2000;' n2 M! p8 R- w4 M. m4 q
   
6. net.trainParam.goal=0.65*10^(-4);
   / y( r4 O% I) i6 w' @' u
7. net=train(net,pn,tn);

复制代码

我只改了这里面的，
4 L; d3 M+ j. Q5 A& ]一、改成单隐含层的，6个节点
* y4 g& Z0 W- _4 C* ^二、训练函数改成梯度下降BP算法 traingd
. G2 Q' d; p0 J/ B. `! [三、迭代次数改成2000
" b$ |4 w: I5 B( Y' U, c7 S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。
: v! N! o! P8 r/ }# v1 A3 f# k3 ~  n! I$ @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