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！！
6 S& U8 [4 N' r4 ^' @$ w/ _
%原始数据输入
% z7 j/ Y! R5 G/ i7 y; aclc
2 i+ A) ^0 `! hsqrs=[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];
% _; m  {3 ?, L) Tsqjdcs=[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];
& L- H9 U- E$ h$ f8 A6 ~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];
: ~1 Q  \1 k8 k) C; v1 y+ y# Qglkyl=[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];
4 l: t# f, l8 G" @3 P7 }+ M" Fp=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

%数据归一化
" L: C& [- ?# z/ l& y/ T[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
7 r& k5 I. m1 M  Ndx=[-1,1;-1,1;-1,1];
8 U' _4 t6 b- J7 r) ^9 G9 T- C
%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
) M/ b, x  x9 _8 V5 S) ?5 tnet.trainParam.Lr=0.05;
! ^! [9 X/ Y1 ~6 P7 Qnet.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
* P: J( d7 j  _9 oan=sim(net,pn);
a=mapminmax('reverse',an,ps2);

, E6 f6 S% W- e2 K/ @, {* h%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,:);
newh=a(2,:);
figure(2);
" B, y5 p/ G) d4 Ssubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
. u) \# w9 q3 Klegend('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');
0 B* L3 n' g8 N( ]8 _7 _4 Y4 Isubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
4 F  @; Q3 I: \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');

' d. T' S: _4 a8 }%新数据仿真
- t8 e7 `% e# B  Jpnew=[73.39,75.55
. p, t- m8 \. Q! F& U: \, Q% H    3.9635,4.0975
    0.9880,1.0268];
# v$ Z9 U7 f/ Upnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
0 N3 W6 g5 W! V) _; canew=mapminmax('reverse',anewn,ps2)
) R6 s5 |8 `6 z+ Q, e
7 Q, q& ^0 J7 i- P8 E

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 ...

: o3 c8 i. Z* J- X% C+ j因为版本更新，Matlab中的Newff命令用法有所改变，原命令为：
5 ?0 |5 E" u& Y7 O8 U0 v7 enet=newff(dx,[3,7,2]{'tansig','tansig','purelin'},'traingdx');
7 S  |" `* n8 N9 I9 K, D我修改为：
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
8 ~& K/ w9 H  M$ T6 J- P6 ?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);
! o' r, r) x4 Canew=postmnmx(anewn,mint,maxt)
0 C& x6 @' w  b7 f5 L% \- W- I' s% Z8 W我修改为：
[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
5 ?! t7 ?* w5 G) X  v+ ua=mapminmax('reverse',an,ps2);
pnewn=mapminmax('apply',pnew,ps1);
8 k$ @- K) J5 }2 j. y- Hanew=mapminmax('reverse',anewn,ps2)
( Y. t% T6 @% w" S, K4 ^
' t' s( t4 d/ q  D8 i- M原程序为：
%原始数据输入
' j2 l3 _# |" `5 Gclc
( ?* X& c3 z1 K$ osqrs=[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];
. b0 Y: j* E- f& s! R3 L" t0 h, 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,...
, i$ o* c* w* \/ v  m    2.5,2.6,2.7,2.85,2.95,3.10];
' c7 U; ]% M6 \& T- ?; Tsqglmj=[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,...
# F8 F2 S1 ~0 {" s8 F5 W/ |0 J    0.56,0.59,0.59,0.67,0.69,0.79];
8 g5 U  O) v' c, Oglkyl=[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];
7 p  a' u0 o9 `5 L2 d+ x- ~/ Xp=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];
8 G+ I/ n# Z6 C$ s
%数据归一化
" I7 k( Z+ {/ j: C[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
dx=[-1,1;-1,1;-1,1];
6 j# V9 f- b  o- D0 \
%BP网络训练
& ?4 P! ]: |  @/ Y& g4 k2 Vnet=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
4 {; r$ s( M( c& Q+ p# ^net.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
7 U- A, a  u& xnet=train(net,pn,tn);

$ l. G7 S* J. P5 B/ D%利用原数据对BP网络仿真
an=sim(net,pn);
- @) x: {5 X- k& a7 ta=postmnmx(an,mint,maxt);
" A1 C9 o" ]5 e+ ^6 K' Y
5 p; T6 B9 @* B%仿真结果与原数据对比测试
x=1990:2009;
6 V) L$ z3 R: |' p( `  ]3 i- Vnewk=a(1,;
newh=a(2,;
figure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
3 x2 J5 L4 a! z7 J/ clegend('网络输出客运量','实际客运量');
6 U% K  j/ R; ], W- J- T) Pxlabel('年份');ylabel('货运量、万人');
/ |4 I3 F7 ~& Z: |& q6 Atitle('运用工具箱客运量学习和测试对比图');
& F' W( ^9 J3 S- b8 n! K9 usubplot(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');
; Z1 L& b5 L, Q5 _title('运用工具箱货运量学习和测试对比图');
9 G/ b9 v7 P. e6 [, D7 t1 i( s
' Y1 ?1 i0 S! J3 u$ H%新数据仿真
; P9 q: [: C5 @* J' Upnew=[73.39,75.55
' X; C3 t- o! o! U" w/ V8 m: e    3.9635,4.0975
# K0 B( g6 ^- }# X) p    0.9880,1.0268];
9 J/ k7 |# x/ `6 v) N2 G  zpnewn=tramnmx(pnew,minp,maxp);
, U7 u9 w  b0 F% u& i1 U; Hanewn=sim(net,pnewn);
anew=postmnmx(anewn,mint,maxt)

修改后程序为：
  W- ~  e" m/ ~; ~%原始数据输入
- |9 n8 ^# ]9 d9 d  V! q* Rclc
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];
; ?9 Q9 Y: }4 K9 s  e' z0 _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,...
& a6 b8 O( i* B8 h. S- m    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,...
) U. u" M: `, M+ t    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 O/ M9 N: a: x. s    22598,25107,33442,36836,40548,42927,43462];
glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
' M+ M; ~1 J8 V, W% y) N    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];
, N4 F  m8 W3 V" z. `* e
0 ]" z) K7 @# j- g3 i& @/ s6 h9 U%数据归一化
[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);

! [: K% J4 U3 v) R7 a. q! {%BP网络训练
8 ]( c3 T6 I5 l- ~4 L" j# C2 ?net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
$ p1 c6 U# l2 @  j5 |net.trainParam.show=1000;
! K! |! c# w% ?0 i9 Z! n: @net.trainParam.Lr=0.05;
; u/ p. f2 f  n; ?9 l4 b( enet.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);

& z( a" ^6 ~2 f# `$ F%利用原数据对BP网络仿真
( R6 Z5 [$ ?4 Z9 ?. R3 {an=sim(net,pn);
a=mapminmax('reverse',an,ps2);

%仿真结果与原数据对比测试
7 G# q% Q9 J, e$ @6 [* dx=1990:2009;
- F" a% ]. f$ mnewk=a(1,;
& W9 N: p- L, W/ C2 o' rnewh=a(2,;
figure(2);
* a2 E( l+ f; `$ l3 dsubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
legend('网络输出客运量','实际客运量');
6 p0 F5 Z8 w* z, T9 v% Fxlabel('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');
! I3 S. h: l* v) \$ C3 Z, `# Usubplot(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');
8 X; D( b) X) @; W0 m. [. jtitle('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
  x9 t/ \2 i3 H3 ^* W2 O    0.9880,1.0268];
) H, [+ |/ i" D0 T7 g$ I3 epnewn=mapminmax('apply',pnew,ps1);
; i. K9 p" a, m$ W2 R( k) @anewn=sim(net,pnewn);
, v( ^; C  z7 z8 P9 {anew=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
   * }% N+ [8 `* H9 n4 ^
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,...- h3 r! K2 u6 u( k* _( u
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   \" z2 n' S' z/ p8 G) Z, ?; @
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 D/ Z  t+ N+ J
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   , a' P& u8 C3 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,...
     O6 o5 N  D4 K- v# Z2 U% V9 P
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   0 Z  C9 Y8 H/ t/ Z& W
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...9 p; J& A# D, d; s4 u. x' B
   
9.      22598,25107,33442,36836,40548,42927,43462];
   1 T\" j, b5 h1 r9 ?! F7 E% m
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...* D( G' q9 B0 W5 [3 X8 M6 h
    
11.      13320,16762,18673,20724,20803,21804];+ U# F7 ]8 t6 O$ i/ M$ Q1 R/ S- [& o
    
12. p=[sqrs;sqjdcs;sqglmj];* i: d  |\" ~( Y0 H1 Z! b
    
13. t=[glkyl;glhyl];
    \" C( D- y2 H; p; s' _/ F
14. %数据归一化; x6 A4 J+ D) }3 w
    
15. [pn,ps1]=mapminmax(p);/ a( Y$ `4 h; U\" O5 `, v& D
    
16. [tn,ps2]=mapminmax(t);
    % w# c2 B* U5 }! l3 @
17. %dx=[-1,1;-1,1;-1,1];7 A6 A! l+ ]' |3 [: ?  X. R. m
    
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]( ^* w0 e8 U. w9 R9 R! A
    
19. 
    : t7 H6 R# C# v. b; u- a' g  k
20. %BP网络训练4 z1 ~7 T+ E7 c& N3 K
    
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]
    - f& |: _8 X+ ~8 }- _3 ]% s\" d
22. net.trainParam.show=1000;; W( z3 ]! s) n# c
    
23. net.trainParam.Lr=0.05;. A9 B* G( D/ J. r) [$ O+ `
    
24. net.trainParam.epochs=50000;$ `* z+ R) s0 e, n
    
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]0 x! _6 _/ h: Y2 N; ?8 h$ O
    
26. net=train(net,pn,tn);
    . _( l6 w5 G- ?2 N4 k. v
27. 
    3 `% k, V) R, y! f8 Q: R
28. %利用原数据对BP网络仿真0 a+ a, M) |2 f6 y0 X\" t
    
29. an=sim(net,pn);& ]8 G1 `/ m: t4 t, H% y
    
30. a=mapminmax('reverse',an,ps2);' Y3 A& W9 m# Z) O4 z3 Q
    
31. 
    ( f6 B$ R9 I% y: m) v! S; X
32. %仿真结果与原数据对比测试 3 \% ?# L1 F% @: r/ Y; @
    
33. x=1990:2009;7 r. H3 t+ N! Y0 K! b& S
    
34. newk=a(1,:);
    # x: ]& r5 W9 ?3 E0 @' U& \
35. newh=a(2,:);4 f7 w' Q5 }$ z/ f! e- ?9 f
    
36. figure(2);) \! {! c3 Z0 A( C8 }
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');& h) k) O  L\" n' I& x
    
38. legend('网络输出客运量','实际客运量');
    / }2 Q6 v& b5 ?0 Q2 p( M
39. xlabel('年份');ylabel('货运量、万人');
    ) V! q- q) n0 I0 D4 W% t
40. title('运用工具箱客运量学习和测试对比图');
    ! r' j! K# g\" @# V. }
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
      z% l  o3 \+ u6 l' ]4 t% O
42. legend('网络输出货运量','实际货运量');! W1 ?+ J/ m! G( D* p# Y
    
43. xlabel('年份');ylabel('货运量、万吨');
    ' ]1 \  ~, W+ p1 l1 \  J6 T! y. v
44. title('运用工具箱货运量学习和测试对比图');
    + m) s8 R& F, U& E
45. \" p  `, h+ w  \0 Q. ^
    
46. %新数据仿真
    ' b  A* Z! T0 z
47. pnew=[73.39,75.55* o/ w  M' X& i\" ]
    
48.      3.9635,4.0975/ g3 b1 Z( x! B2 t- C- D8 z
    
49.      0.9880,1.0268];
    + `8 C/ I: s5 m& u; G* ]6 ^
50. pnewn=mapminmax('apply',pnew,ps1);% g2 K9 g& H3 i- g9 M% ^
    
51. anewn=sim(net,pnewn);
    3 o$ m+ B  X: J. d: P. i
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. clc0 m$ B( W1 l) X4 `: ]\" S
   
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,...
   $ w' Y5 ^0 ~& A7 K( ^7 K2 r
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   2 A\" o$ r3 W, W
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 ?3 Q! C2 v* r. y) V; O( p; }
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];5 U\" B/ |2 l, }  m6 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,...
   1 p- O( b& O$ t& t0 \' }
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   2 b  ~1 N$ c1 N0 K, Z; Y$ c
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...( E/ D' R% P+ ~0 z/ B3 ?
   
9.      22598,25107,33442,36836,40548,42927,43462];\" |! ]% }  f( O7 ?8 d) a
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...) M) X) ^* c$ H& F# O+ O: J( ~7 f
    
11.      13320,16762,18673,20724,20803,21804];
    0 @+ t6 C/ ?; s7 U( M
12. p=[sqrs;sqjdcs;sqglmj];
    1 @# I3 |. b2 R  K) P' u
13. t=[glkyl;glhyl];; f/ @2 X& ^0 R
    
14. %数据归一化
    4 J9 M. e. H+ H. z0 w
15. [pn,ps1]=mapminmax(p);
      e3 g! l2 s$ e+ f* G  _
16. [tn,ps2]=mapminmax(t);/ b) F/ G1 r/ Q; w$ G9 F) F2 d& N
    
17. %dx=[-1,1;-1,1;-1,1];
    : x2 k( Q- C- D  A) n8 m9 w
18. p0=minmax(pn);t0=minmax(tn);
    ) A' M. ?/ J' y4 L7 D  S9 Q- q
19. , t3 m3 g7 O\" a0 k0 a, U
    
20. %BP网络训练
    # k+ q+ Z+ i/ _4 ~$ R( L$ y. q% g
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); / x6 ~: c4 U. A\" j
    
22. net.trainParam.show=1000;2 w( ^7 C) M. K, o, U8 X, h
    
23. net.trainParam.Lr=0.05;, ?0 D6 K, F# h, u  \' s
    
24. net.trainParam.epochs=50000;/ R* s2 o. F* q& c( z9 S
    
25. net.trainParam.goal=0.65*10^(-5);
    2 ^( L; Z. g8 V; |5 \  X
26. net=train(net,pn,tn);  m% y  V3 k3 C8 Q  C7 \; x$ F- k
    
27. . \( m- \, ^- g: u. A
    
28. %利用原数据对BP网络仿真
    & r' o+ i\" _2 x  V
29. an=sim(net,pn);
    ) o. ^4 O  e6 S5 ]
30. a=mapminmax('reverse',an,ps2);
    * m9 }\" }. u9 k. U* ^) ~
31. ! o, a. U# y5 |% J
    
32. %仿真结果与原数据对比测试 ; D8 h* F# T/ K3 M) }4 G
    
33. x=1990:2009;
    \" ]. T+ ^0 c( A' S  ^+ Y
34. newk=a(1,:);
    : e7 n0 }7 o  |; [  Y( g
35. newh=a(2,:);
    8 Z6 @) p\" M9 D' T, c; h
36. figure(2);
    , R4 g8 C& R/ t$ R! p
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    0 X/ d/ H+ p/ V! O6 o9 X! M& n
38. legend('网络输出客运量','实际客运量');1 d; O' y' V; \0 j: [2 y
    
39. xlabel('年份');ylabel('货运量、万人');5 i/ f6 i. I3 X2 [4 A/ q
    
40. title('运用工具箱客运量学习和测试对比图');
    , W$ |0 L/ g8 }% w
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');4 M- O9 ~  W! [; L
    
42. legend('网络输出货运量','实际货运量');/ l  i% ?! h: r) s% u
    
43. xlabel('年份');ylabel('货运量、万吨');% G' v& k$ O$ j8 e
    
44. title('运用工具箱货运量学习和测试对比图');2 S4 b* g9 L+ j* ?' m) `' p# `
    
45. 6 o7 K, C& E. @0 S& H$ M, Z# o
    
46. %新数据仿真9 N3 u& K! s% ^0 V, a& Y
    
47. pnew=[73.39,75.55
      a+ Z' J% V3 a7 X' R5 H. F
48.      3.9635,4.0975, o2 A1 A2 L) X$ V. Q2 M3 k
    
49.      0.9880,1.0268];
    % g' ?. N) x# F4 o0 n% x
50. pnewn=mapminmax('apply',pnew,ps1);  u4 v; V* v8 Y2 _
    
51. anewn=sim(net,pnewn);3 Z8 I0 J/ n9 m  c% o0 Y
    
52. anew=mapminmax('reverse',anewn,ps2)

复制代码

T-Eric

且生 发表于 2014-7-21 18:22
7 s/ L/ a7 ?- z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
   . u  K: T4 I. O0 [* N; A6 `, M
2. net=newff(p0,t0,6,{'tansig'},'traingd');
   # ?0 Y! }! I( H: L5 O9 ?' ~9 I% c7 k
3. net.trainParam.show=1000;% ]  a. _4 l: l7 R& ?
   
4. net.trainParam.Lr=0.05;
   . }, T# `8 v5 o* G! M1 z6 `6 @
5. net.trainParam.epochs=2000;7 _( M2 {+ B% F* Q9 u, V
   
6. net.trainParam.goal=0.65*10^(-4);' f6 ^( H! w7 a
   
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
1 e4 p7 G4 e8 }! @上面的参数是自己试的，我也不知道为什么。
  T  ~  x/ [: x; {1 \) F9 |5 Y/ {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