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！！
: b3 o' D: a& o* C: i) \& z6 [
%原始数据输入
' o3 C: _6 o$ D" U* c3 Q- mclc
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,...
! N1 j6 c0 Z3 [! }; G/ D% i1 M    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
3 r+ b+ Y; m5 h1 G8 }8 esqjdcs=[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];
& o- T' ~* \" psqglmj=[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];
* w  h, i3 ]% u. F/ Y2 E; V& ?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,...
5 z  _; w9 o2 Z' E/ c9 a    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
/ \/ b" h- m2 q8 Q% V2 ct=[glkyl;glhyl];

$ g, }; v4 c; i9 I2 e6 t  H5 H%数据归一化
[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
dx=[-1,1;-1,1;-1,1];
- m/ \1 w  D! M  I% T, n
%BP网络训练
$ B' Y% s6 n8 U+ b: u+ \! Enet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
net.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
: F3 _0 r1 }( Q$ F2 B' _net.trainParam.goal=0.65*10^(-3);
0 T* b9 a# \2 [, hnet=train(net,pn,tn);

  k) h% R4 _* j' p%利用原数据对BP网络仿真
5 U0 G: J2 h) |an=sim(net,pn);
/ r) D/ F, z- {/ O8 ua=mapminmax('reverse',an,ps2);
+ l9 P% v3 B+ n7 H/ M# B1 f
. {% y0 M) V+ N1 m+ |4 O# I' S%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,:);
. J1 T2 ~; B) z( v/ \newh=a(2,:);
figure(2);
, d$ J) w' r+ H5 W# f& [# Csubplot(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 M9 |4 A' M3 \9 H$ z5 cxlabel('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--+');
  b8 a) k! y. h( |9 ?6 Mlegend('网络输出货运量','实际货运量');
4 r1 }: F1 B# rxlabel('年份');ylabel('货运量、万吨');
" q; A  G3 J' [9 b7 Ntitle('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];
pnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
: ^$ b3 }4 D& j! L8 a# yanew=mapminmax('reverse',anewn,ps2)
) l1 F* Q  M, O+ ?5 V. \
" s6 l+ ]' b7 L" f) 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
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 ...

4 r& ?1 y3 Y. Y$ f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：
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：
[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
  R  U# l. k; _a=postmnmx(an,mint,maxt);
) ?# `3 L" n9 P7 ?7 S6 Opnewn=tramnmx(pnew,minp,maxp);
" H& m9 \- Y, tanew=postmnmx(anewn,mint,maxt)
: ]& e$ b3 ]% K6 o+ a  h. ~我修改为：
2 [2 y6 C( ?6 R/ k[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
a=mapminmax('reverse',an,ps2);
pnewn=mapminmax('apply',pnew,ps1);
anew=mapminmax('reverse',anewn,ps2)
: K1 e& ?' B8 X
0 ~+ G$ n5 w! m" I" e5 z原程序为：
8 ^2 Z' h: x8 @& E0 S$ [/ s%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];
' x0 _1 D( ?. B9 l+ k9 f$ G# |8 `, B! r) Esqjdcs=[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,...
' k6 c: S$ l; D    2.5,2.6,2.7,2.85,2.95,3.10];
! J" g! J& d0 X/ N; T2 g# \1 Usqglmj=[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];
# l) z/ Q! |  M8 o. g1 g8 Q: pglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
' k+ W0 }5 _' ^2 W. t7 e    22598,25107,33442,36836,40548,42927,43462];
3 L& r' x1 R' {7 ^  Vglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
$ l4 W( n) j; e) ]t=[glkyl;glhyl];
$ ~8 l1 [& g& K: U% B
%数据归一化
) b( z6 [$ f1 D# |" K[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
( N3 t  ^) F6 }5 z/ _) `/ hdx=[-1,1;-1,1;-1,1];

" t  {) h( t% e3 g%BP网络训练
net=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
5 X$ n5 _: K" Q8 `5 L" O# }net.trainParam.show=1000;
8 E: I3 W, f* a+ V. t# _. Knet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
6 E* G) C5 G4 \* Knet.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);

%利用原数据对BP网络仿真
' b  V! A4 y) s5 _an=sim(net,pn);
& t/ f4 ~9 B. E% Ua=postmnmx(an,mint,maxt);

%仿真结果与原数据对比测试
3 k2 ~" ~- {1 t8 G$ Zx=1990:2009;
newk=a(1,;
newh=a(2,;
figure(2);
4 V. ^5 K) R3 D' p8 u/ Wsubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
9 U. ^, |2 j. q1 Klegend('网络输出客运量','实际客运量');
4 H& r  ]/ |( n% c" j2 rxlabel('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--+');
7 Y3 e& O  m2 W/ Y1 Dlegend('网络输出货运量','实际货运量');
xlabel('年份');ylabel('货运量、万吨');
8 f8 v- V- H$ A! Ititle('运用工具箱货运量学习和测试对比图');

$ g& O9 n( |$ L2 U/ r%新数据仿真
7 Q! L( e0 V, }. f  P. Spnew=[73.39,75.55
9 u% L1 S- [0 O! b    3.9635,4.0975
    0.9880,1.0268];
5 r& F8 t/ n: v- I- w1 Hpnewn=tramnmx(pnew,minp,maxp);
anewn=sim(net,pnewn);
anew=postmnmx(anewn,mint,maxt)
/ g& x4 t& ?. L8 b0 q4 z! M
* a; I8 Q0 U7 e9 }修改后程序为：
; ^1 x) D0 U: T  G%原始数据输入
. `# ?0 t$ G+ v* j) v& oclc
- H9 G1 g2 M0 m7 n( M0 ~9 ]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,...
    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
4 l$ i; |7 ?9 S6 ~+ }/ Y3 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];
# W5 G  k" A' Y6 }6 msqglmj=[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,...
+ m3 Y2 k2 w- p! H8 C) Q    22598,25107,33442,36836,40548,42927,43462];
, e9 z* h# T. J5 V" @9 Nglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    13320,16762,18673,20724,20803,21804];
6 B- s5 R! c+ {; ~0 e0 I/ k  Jp=[sqrs;sqjdcs;sqglmj];
! Y4 W2 ]2 e; H* A, n$ St=[glkyl;glhyl];
0 Q* X$ s2 C+ ^, s- z! i* v
%数据归一化
[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);

. p* w$ k8 w+ t%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
( F% |: _" m$ Z5 Dnet.trainParam.Lr=0.05;
" v( I! X/ ]9 \3 U! tnet.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
" X( x' P' G" E9 inet=train(net,pn,tn);

. H( x9 J( E" ?* H%利用原数据对BP网络仿真
an=sim(net,pn);
& y$ }3 x$ f+ Q9 K" {- h( Ea=mapminmax('reverse',an,ps2);

) L* r6 t8 w; ^%仿真结果与原数据对比测试
x=1990:2009;
7 V$ u% d. y& e* l- ]! \3 @9 Snewk=a(1,;
! o' f+ i( Q% Enewh=a(2,;
0 z) k1 @5 e/ V$ ]5 hfigure(2);
( j' n3 X; P- t" asubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
6 `6 G# @  z$ t9 R& M/ }$ elegend('网络输出客运量','实际客运量');
3 D! O  {, `  k$ g# l* y2 Q- K; B7 Jxlabel('年份');ylabel('货运量、万人');
# `7 J4 m8 h) _# \+ ftitle('运用工具箱客运量学习和测试对比图');
  o0 r1 l, j: ssubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
: V  v- X9 L/ W3 B: C5 K, K* F( N$ ylegend('网络输出货运量','实际货运量');
- U+ x; l9 M3 b6 x# C: oxlabel('年份');ylabel('货运量、万吨');
  a  j& h3 ^6 L+ D. ntitle('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
. h5 I( m- o$ Z9 x) r" _4 V1 ]" u    3.9635,4.0975
" x9 Q! _" x; Z# U1 P  Y    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
+ T1 L& ~& ]5 h( H0 ganew=mapminmax('reverse',anewn,ps2)
3 y; n- K# D* b# M3 A（修改的地方用颜色标记了）
. Y, C+ f  p3 M. J; o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: r; \5 Z5 Q! K( Y+ x
   
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 u2 _6 f# F% s) S1 a
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];& V% `  `. _6 n\" X& |\" ]5 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,...
   2 ]& I& z  p( \* \2 t* c- W# r7 g* C2 r
5.     2.5,2.6,2.7,2.85,2.95,3.10];' d, K2 q) a( c7 L9 T0 c: n3 F
   
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,...+ t+ N4 A# n/ R* s# o& ?& a# z
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   2 ^! w! D- @& y$ r5 Z4 a; s* o
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   ( R% `9 W: S\" B1 [; u
9.      22598,25107,33442,36836,40548,42927,43462];
   ' B/ s- F  T1 W- J$ i& |7 D. r
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...6 |\" e8 G6 O0 Z2 m7 ?, \4 a\" C  ?
    
11.      13320,16762,18673,20724,20803,21804];$ z, W6 c* o9 @! {2 Y
    
12. p=[sqrs;sqjdcs;sqglmj];9 M9 `\" I* c- O) ]# J
    
13. t=[glkyl;glhyl];+ V3 j+ T8 h) Q5 h/ B
    
14. %数据归一化5 s% U+ `\" t. g1 x5 i
    
15. [pn,ps1]=mapminmax(p);9 H4 ~* b% |& W$ k
    
16. [tn,ps2]=mapminmax(t);
    $ X1 a- P$ u* B; P4 D7 t3 t
17. %dx=[-1,1;-1,1;-1,1];
    % c; M- _3 U5 }\" W* t
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]
    - d( W' n8 D! L. e
19. 
    \" }5 s+ ]* X; o5 f7 p! N
20. %BP网络训练+ T) d8 X( f4 V+ }. P2 P* \
    
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color], w! ^; X' ~\" y5 D5 H1 k) Q7 Y, V
    
22. net.trainParam.show=1000;( r# X1 H+ S4 g9 ]( t& p- ~* N0 V% G4 h
    
23. net.trainParam.Lr=0.05;
    4 A5 j4 R; V4 C\" O( t& j2 R
24. net.trainParam.epochs=50000;
    2 [, v; W. ?& p
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]
    . A: J) k0 J( V% U) q# I/ D1 ~
26. net=train(net,pn,tn);: l# |3 `\" Z( j) x# J
    
27. & R) B4 |; A5 L8 }
    
28. %利用原数据对BP网络仿真
    ; W& i' ?+ S8 `+ n
29. an=sim(net,pn);' r5 U/ @# o8 p
    
30. a=mapminmax('reverse',an,ps2);
    3 g5 r* f! U; B' s4 ?5 \8 \' t! N, G
31. : }2 r4 r& p  W2 q
    
32. %仿真结果与原数据对比测试 ( n' B3 ~. R0 l& |
    
33. x=1990:2009;2 _5 ]. I' V! E$ T! Q4 w. y
    
34. newk=a(1,:);9 q' O0 l- W) E6 K; W3 a
    
35. newh=a(2,:);
    . L+ G. n( {! _% f9 w( i9 s, e
36. figure(2);  N; W1 M: I) L; @9 M
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    % V4 G+ U\" q5 X& M  l* g5 k
38. legend('网络输出客运量','实际客运量');
    * h. n6 q7 P4 `6 }. |
39. xlabel('年份');ylabel('货运量、万人');
    7 a7 n9 h5 S4 K: R* W0 x; B
40. title('运用工具箱客运量学习和测试对比图');
    9 V* [2 W2 |% F/ A1 l0 A
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    ; t\" G$ R% |\" @
42. legend('网络输出货运量','实际货运量');$ x  U* |. a! b\" @
    
43. xlabel('年份');ylabel('货运量、万吨');
    % N\" Y& Q7 {/ u4 \) i! E, k
44. title('运用工具箱货运量学习和测试对比图');% k; ?! y( F% V# `9 D0 n
    
45. 5 J! `2 _3 _6 l& [% n4 M
    
46. %新数据仿真
    % t4 W- H5 N( u' C7 W* H
47. pnew=[73.39,75.550 A2 K/ ?( d: W
    
48.      3.9635,4.09751 }1 r* ^' x( t9 a( `, N/ n- \
    
49.      0.9880,1.0268];! c! i4 j7 `$ w( O0 @
    
50. pnewn=mapminmax('apply',pnew,ps1);
    3 x+ ?, f9 P' P$ g. b
51. anewn=sim(net,pnewn);
    # k. y4 ]; V6 z5 }1 N
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& ~% i4 `$ ?0 h/ f5 w
   
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 I  T3 l6 z3 G$ I& |
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];# r0 \0 f' E1 P, X
   
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,...
   : y. G$ f1 F- N7 ?$ q7 i. C
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   ! F0 j+ C5 B( [\" 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,...
   6 n% P( Y) E\" z
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   ) m: q2 U% E/ b9 \/ ?+ k& t+ j
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...1 e1 E- ~5 K: c& f5 Y9 y) @
   
9.      22598,25107,33442,36836,40548,42927,43462];
   ( ]- M2 q- Z3 c% w
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    * j9 R+ B0 C2 `; [: O; t: ?* z; W- O
11.      13320,16762,18673,20724,20803,21804];
    $ ^2 t' _; C/ [' t
12. p=[sqrs;sqjdcs;sqglmj];+ W9 y7 U$ @5 D1 e6 _3 P
    
13. t=[glkyl;glhyl];
    8 t5 p5 k* u+ F! N) S% b1 Y; S
14. %数据归一化
    - J7 y# Y( L0 ^: b
15. [pn,ps1]=mapminmax(p);- D% F/ v  t3 |) |\" ?
    
16. [tn,ps2]=mapminmax(t);$ m- C. L+ Q7 T! P* A9 r
    
17. %dx=[-1,1;-1,1;-1,1];) v) Z% M$ k# k) \+ A
    
18. p0=minmax(pn);t0=minmax(tn);; Q! O: S; b) z4 ^- C
    
19. : O% W\" u2 J# Q1 y: ^: e
    
20. %BP网络训练
    8 g; t4 O1 l0 v& [% [
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx');   C, f* w/ p1 H9 z) p9 a  {
    
22. net.trainParam.show=1000;
    1 S1 z! @2 b& h4 q; R: |7 x0 Z! ^
23. net.trainParam.Lr=0.05;
    5 |: `! h4 a4 y( L9 E4 L2 M; \
24. net.trainParam.epochs=50000;, ~% v6 C+ g% O* {
    
25. net.trainParam.goal=0.65*10^(-5);
    + n/ H0 c9 `6 y+ F8 o3 u
26. net=train(net,pn,tn);$ Y  {! E5 ]6 o* Y- n
    
27. 
    & e( T' _# X$ C\" @
28. %利用原数据对BP网络仿真
    \" {0 q8 r! [. R' Q' ~  D- i/ b
29. an=sim(net,pn);
    5 D; H8 ]; \; W- u. j4 g: s' G
30. a=mapminmax('reverse',an,ps2);; q6 U9 I3 t3 z4 }% M; r
    
31. 3 W; f3 e* c$ k8 D+ s9 a. s+ H
    
32. %仿真结果与原数据对比测试
    ' U3 C/ A+ V3 [
33. x=1990:2009;2 {& U' i2 R; p0 H2 G
    
34. newk=a(1,:);
    . r# h& r9 X& Z% q' K3 x
35. newh=a(2,:);# W8 X! n+ R- ?( i. K9 j
    
36. figure(2);
    + t' h  i8 Y4 a$ ^
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');% y% r6 K' |% o; i) Q& l
    
38. legend('网络输出客运量','实际客运量');
    0 K& N' C# Y, }/ U8 d1 o
39. xlabel('年份');ylabel('货运量、万人');
    6 R3 j\" ]/ D% u$ v2 k2 m- p' t
40. title('运用工具箱客运量学习和测试对比图');- i) U! |8 u( [$ h  [
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    5 e* u. Q0 l! R5 i
42. legend('网络输出货运量','实际货运量');
      ~6 |9 k, x% E' x
43. xlabel('年份');ylabel('货运量、万吨');
    1 X: Q( P* z9 a, ]$ L
44. title('运用工具箱货运量学习和测试对比图');$ O3 z+ }* I8 h) N/ w2 Z# ~' L
    
45. ; F% e8 d0 M( }) e4 d* F/ e
    
46. %新数据仿真8 D7 Y, ^# a) z/ t  j% d* Y! D
    
47. pnew=[73.39,75.55' @5 d0 D3 A3 s9 D; A
    
48.      3.9635,4.0975\" f1 x  T; ?0 M\" W8 Q* S
    
49.      0.9880,1.0268];1 k2 M, z: x+ }) H; N+ _9 F4 @
    
50. pnewn=mapminmax('apply',pnew,ps1);
    / z0 m7 r% [0 ]4 L5 f\" y
51. anewn=sim(net,pnewn);$ _; _$ X' Z\" l
    
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

  J! ]( a' \" t! y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
   9 o: B0 m' x/ K/ r, H\" M
2. net=newff(p0,t0,6,{'tansig'},'traingd'); \" _4 z$ N! U/ a
   
3. net.trainParam.show=1000;7 J3 O+ h' E# J2 y) z) v8 m
   
4. net.trainParam.Lr=0.05;
   8 C  F+ X: r' A* C
5. net.trainParam.epochs=2000;/ R: K# j( u/ B3 s1 Y
   
6. net.trainParam.goal=0.65*10^(-4);: H) s3 V- u$ Y' t, i  H. Q
   
7. net=train(net,pn,tn);

复制代码

我只改了这里面的，
" V% N0 J. \7 ]" R. p  z一、改成单隐含层的，6个节点
0 w1 V0 Y& v) C0 M二、训练函数改成梯度下降BP算法 traingd
7 T) X6 n' b, r, _4 ?) i! r# [1 E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。
" o8 U7 o* C" a6 {; U$ V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。
' s1 j/ h+ r# T9 Z3 w$ _) k2 W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