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！！

/ J  B$ ]. o7 t! _: r%原始数据输入
) L* J* H) L  @3 J+ gclc
, ^$ H; g5 i  R) V0 J  w6 usqrs=[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];
* f4 a5 D" W- y- H! g6 p" Psqjdcs=[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];
4 X# }6 G9 |; `! U  X+ ksqglmj=[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 L8 f5 J$ ?$ j) ]+ i    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,...
6 w1 _/ y7 o4 J0 A# |( u' B# U8 s    22598,25107,33442,36836,40548,42927,43462];
glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
9 N# ]+ \2 _4 d4 e6 G/ b- d) B( j    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

%数据归一化
[pn,ps1]=mapminmax(p);
& W, z$ ]0 I1 o9 q[tn,ps2]=mapminmax(t);
dx=[-1,1;-1,1;-1,1];
4 w* m9 F- g/ M$ H! O) ]# I
%BP网络训练
$ _# }- ]) g7 u+ X5 [" A, _- hnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
; Y" |& {$ v1 N  m% W% ynet.trainParam.show=1000;
* ^6 u, z! o: }5 u( Vnet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
0 z+ F- i) Y6 u& d; cnet.trainParam.goal=0.65*10^(-3);
! O, G/ e: B, Rnet=train(net,pn,tn);
2 _: ?9 |, q. r2 n3 |! h
8 I. r. Z- P7 T8 |%利用原数据对BP网络仿真
. [5 C; H6 I* V/ h( J* g* i: `: Ean=sim(net,pn);
5 ]2 J8 ^6 @8 n  Q1 sa=mapminmax('reverse',an,ps2);
: X% a3 d/ z% P# l3 W9 C
%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,:);
% `: \- Z/ l- m5 a9 Knewh=a(2,:);
" X: o% s: W8 Lfigure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
$ p" a( _7 U6 N! B! }4 m6 Zlegend('网络输出客运量','实际客运量');
xlabel('年份');ylabel('货运量、万人');
  I$ J8 Z/ o9 N6 e- ptitle('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--+');
" T& c% D3 O6 @, t" Y) y. ^1 flegend('网络输出货运量','实际货运量');
+ e2 |7 l  \6 c$ o9 dxlabel('年份');ylabel('货运量、万吨');
, y1 r% C7 {% h. j. D* f7 Otitle('运用工具箱货运量学习和测试对比图');

: D' }; H5 C8 _%新数据仿真
pnew=[73.39,75.55
  X$ Z0 L. \5 G. \, [9 M3 u9 D    3.9635,4.0975
0 p' g0 y1 X8 ?5 ~1 F    0.9880,1.0268];
4 B) `: r( m) a; B7 mpnewn=mapminmax('apply',pnew,ps1);
; h* T- {8 K5 y. @anewn=sim(net,pnewn);
anew=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
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 ...

因为版本更新，Matlab中的Newff命令用法有所改变，原命令为：
/ Y3 i2 K: r3 J3 P5 O7 o1 |! W1 p0 onet=newff(dx,[3,7,2]{'tansig','tansig','purelin'},'traingdx');
我修改为：
; _& W0 v& {. c6 V1 T, F1 [. e: M1 xnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
2 V. L  W/ ^' R' V/ X1 \) Q# s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：
9 @: j3 n7 E9 C7 q2 n/ O3 x[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
- t7 d$ X* \) E  ]% h5 ^* [; L8 _# @a=postmnmx(an,mint,maxt);
pnewn=tramnmx(pnew,minp,maxp);
7 O& k( @5 J0 D( Danew=postmnmx(anewn,mint,maxt)
我修改为：
[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
a=mapminmax('reverse',an,ps2);
  ~9 h# e. x' r% Z1 H& n7 n2 Npnewn=mapminmax('apply',pnew,ps1);
' Q3 U1 Y/ q9 u6 Banew=mapminmax('reverse',anewn,ps2)

原程序为：
' G* W3 o$ V- p$ L%原始数据输入
: W6 z6 G6 @' Q$ s/ b8 E; Z; Tclc
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,...
    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,...
* f: J- `, d+ }% X# q8 \; C! U    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];
" ^6 x0 H6 Y9 g: B/ rglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
9 L. j0 E5 c: G( {& D4 r0 U% F, @    13320,16762,18673,20724,20803,21804];
) b% r& d$ K$ L' R# L; Sp=[sqrs;sqjdcs;sqglmj];
) A  e- r. U# I5 i3 et=[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
+ W3 f* \7 {. xnet=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
( P8 c6 v5 u. U+ inet.trainParam.show=1000;
net.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
' ]+ j8 e8 n, ^4 t3 inet.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);

%利用原数据对BP网络仿真
* l. Z1 A0 y! O4 F, {* c; dan=sim(net,pn);
- P/ L$ U$ L4 @5 C1 t' j" \1 b/ a; ?5 Ia=postmnmx(an,mint,maxt);

%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,;
newh=a(2,;
0 q. V: f1 B2 ]+ Wfigure(2);
5 [9 _5 r# H" [+ A7 q; c% F' 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');
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');
. A: L8 Y% V0 T2 qxlabel('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');
, J  V- ^( J' m8 P
%新数据仿真
& K( m+ ], Y' N9 ?) i9 Qpnew=[73.39,75.55
    3.9635,4.0975
    0.9880,1.0268];
pnewn=tramnmx(pnew,minp,maxp);
anewn=sim(net,pnewn);
  w: _' c: I1 a, [, \% u' hanew=postmnmx(anewn,mint,maxt)

* h9 K; O- m1 l修改后程序为：
%原始数据输入
3 W. }9 {( Q" h1 ~1 d. s# k& }2 Pclc
+ k4 G) e1 H9 u; S% t) d; B3 Rsqrs=[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];
1 {0 k, q8 e4 W7 O# t1 k- ksqjdcs=[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,...
( a+ F5 Y8 \# J1 ~    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];
7 V- S$ w1 U8 S. J8 oglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
    22598,25107,33442,36836,40548,42927,43462];
3 ~$ J% K0 N/ [! a5 Y4 U. u/ ]glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    13320,16762,18673,20724,20803,21804];
7 D" V, V7 j1 I8 y5 K; f( T# a) x: fp=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

%数据归一化
[pn,ps1]=mapminmax(p);
3 C  I: Y, {4 A1 l! s[tn,ps2]=mapminmax(t);
& I+ l" H/ i) P
6 a. h$ e, ^+ `) z%BP网络训练
, U2 p# B0 F) x4 ynet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
7 U+ v- ^. u/ U4 @net.trainParam.show=1000;
net.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
* L3 m: T7 k1 {% ~8 j* G! Onet.trainParam.goal=0.65*10^(-3);
( d% C/ l0 o" u$ ?# e" B5 Anet=train(net,pn,tn);

  Z5 s. [6 O3 h%利用原数据对BP网络仿真
  I6 A' _; n7 `/ O) ~+ p0 f1 Xan=sim(net,pn);
% B+ u1 k" l8 B% B- R# p4 D3 d0 Ka=mapminmax('reverse',an,ps2);

%仿真结果与原数据对比测试
! {; {: M, f: h: u8 ex=1990:2009;
& W0 ^: @9 l  E5 X, Jnewk=a(1,;
% l& _9 @& |8 p6 c, G$ Ynewh=a(2,;
2 z3 W8 M% b# t, z) N) qfigure(2);
+ [4 g* L: `' l* S) csubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
4 u* ~: D9 C8 |+ ^& r7 qlegend('网络输出客运量','实际客运量');
xlabel('年份');ylabel('货运量、万人');
7 |! S  C) u. R$ g0 ktitle('运用工具箱客运量学习和测试对比图');
/ J$ N5 D$ [: u, B3 fsubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
legend('网络输出货运量','实际货运量');
! Z1 i2 j! G. a! b$ _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');
9 z( K/ j  x4 f' ~; h6 O
%新数据仿真
pnew=[73.39,75.55
( A+ s" k7 X, \$ R9 G' x  h" o    3.9635,4.0975
: M$ y4 X1 t+ A& |1 O' j1 H    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
* X8 t0 `5 O* P* k6 ^% {anewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
" a! ]/ B* ]( |  \! d/ Z（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
   5 @8 h' M, N$ T: l9 s9 u
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,...& J& r; P7 @2 t- O7 s; A9 c! l: q- e
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];3 o; |5 r9 D! f: b6 O& o
   
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,...- d9 k6 ?3 x: o1 s\" Q4 _- f8 g+ x
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];7 C. A$ I6 w9 T3 X% Z
   
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,...
   ( C. c3 Y& ?) q. l7 ^' ?
7.     0.56,0.59,0.59,0.67,0.69,0.79];$ Y. B0 W6 E: _$ n
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...7 s( C  I* E' w/ A- _
   
9.      22598,25107,33442,36836,40548,42927,43462];( H: C# q( G1 q4 {, K+ [, v& s
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    2 G9 e# i/ w% T  y
11.      13320,16762,18673,20724,20803,21804];
    - f, v$ {' m/ |4 T4 w, R
12. p=[sqrs;sqjdcs;sqglmj];
    1 S+ J3 v+ M4 I( T
13. t=[glkyl;glhyl];7 ~! j8 [! W2 z/ c0 J
    
14. %数据归一化
    # T) ?2 ]6 s! L3 w; l. L3 H
15. [pn,ps1]=mapminmax(p);
    * V5 Q8 i# a8 y  y' \4 w+ K
16. [tn,ps2]=mapminmax(t);2 t2 w; D2 ^  e
    
17. %dx=[-1,1;-1,1;-1,1];
    # Z7 J. u# |; n4 }' i
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]( l8 r' L' \- G
    
19. 
    / g( S; S0 M1 I5 O9 x# j\" s, Z
20. %BP网络训练4 r) w/ E6 ^9 R  H& Z8 k  v
    
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]
    3 C* n  `# c& _) D% A- ?, C- b, x
22. net.trainParam.show=1000;
    7 L& v+ V- t2 y6 P9 e* {
23. net.trainParam.Lr=0.05;% ]; Z# p- g* O) {; k
    
24. net.trainParam.epochs=50000;
    , h; Z5 Q' v4 u# E+ a/ d
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]
    $ J2 \! a' n4 a: |, c3 k; O5 x\" G' M, Z: R
26. net=train(net,pn,tn);; a3 ?  B; ?% y4 g7 t
    
27. ; @# e% f5 O! @- l1 l) N+ Z9 h  h
    
28. %利用原数据对BP网络仿真: a) q0 ]* b) S
    
29. an=sim(net,pn);
    $ s, o, o, v, d* j
30. a=mapminmax('reverse',an,ps2);
    - Z' z, ~& y8 f9 V3 _; ^2 F
31. 8 g1 R\" D* ?) q* _% T$ u
    
32. %仿真结果与原数据对比测试 5 m9 h: @! f/ T8 h
    
33. x=1990:2009;
    1 }# V, R! |& l+ J
34. newk=a(1,:);\" j  x' u- l& [0 e' E
    
35. newh=a(2,:);4 G$ v# f5 F* ]+ }# C6 z8 e2 b
    
36. figure(2);
    9 y5 e, w: C5 Z8 i
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    + {. R- o0 B\" r$ F
38. legend('网络输出客运量','实际客运量');2 B\" ^: P% k) s% a5 w/ {1 M
    
39. xlabel('年份');ylabel('货运量、万人');
    3 K- Q4 [0 w/ f+ Z  g, F
40. title('运用工具箱客运量学习和测试对比图');
    7 l& y! A  v1 i; L\" q1 v9 z' \
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    5 r5 Y  Y2 c0 w( u7 U# k$ n- \; n8 \
42. legend('网络输出货运量','实际货运量');
    & P7 t4 w) |\" h# \( I\" ^0 S
43. xlabel('年份');ylabel('货运量、万吨');
    : z% Q( A. X- N/ I5 m
44. title('运用工具箱货运量学习和测试对比图');0 r1 U2 t8 z  B0 l3 x3 W' L$ k
    
45. 
    - `3 P4 w\" r) A2 `
46. %新数据仿真) [, x. |' b# F+ ?- b8 z
    
47. pnew=[73.39,75.55+ }* _; ^( B0 y- _$ x6 p
    
48.      3.9635,4.0975
    3 x; Z\" g% Z5 O6 M
49.      0.9880,1.0268];
    * y! ^5 l% ~# k, T) Y- @% x
50. pnewn=mapminmax('apply',pnew,ps1);
    3 p* }3 _6 J, d) Y
51. anewn=sim(net,pnewn);
    - X9 a( h) E3 F! C2 v! z
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' e7 d2 @$ Z+ h+ `, h; F
   
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 ]& g! ]; Z
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];% m# O0 g+ }( k7 x) J  p( e
   
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,...
   - L1 B! m# n! H- V' _8 `& _; k
5.     2.5,2.6,2.7,2.85,2.95,3.10];, j' Y+ h; X6 H2 S0 E! M  L0 s\" M! G
   
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,...
   7 ^% V1 b/ B! v& \5 r
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   9 ]# H5 S, S6 n8 R
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...; ^- o! Y: J( T0 f. `# }
   
9.      22598,25107,33442,36836,40548,42927,43462];
   & i# {! f' ~' v7 i! w( p8 o
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...5 w0 Z1 B9 R5 h  G& K1 u
    
11.      13320,16762,18673,20724,20803,21804];
      @1 D1 ]! ^( |: S
12. p=[sqrs;sqjdcs;sqglmj];8 r: ?- y* x. _/ o3 F4 }6 U
    
13. t=[glkyl;glhyl];( J7 r) H; G; Q9 ?4 G
    
14. %数据归一化# @! Z8 a- I1 O8 \; ?9 K. s% O
    
15. [pn,ps1]=mapminmax(p);
    ! }7 B' Q/ i* g. T: W
16. [tn,ps2]=mapminmax(t);
    / z) M% `0 @9 Y) H, i; z) a
17. %dx=[-1,1;-1,1;-1,1];9 P7 Z% W* s' R
    
18. p0=minmax(pn);t0=minmax(tn);
      G4 ~9 K$ l9 d* ]1 D% B5 e' B
19. 
    0 H0 m. f: E# F* @0 P+ l
20. %BP网络训练7 E4 |/ b2 S, w1 R. o6 `+ N
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); # M. w: R: B2 U, D8 K) i4 T
    
22. net.trainParam.show=1000;
    6 s/ _. Q3 B' B8 c( P0 v
23. net.trainParam.Lr=0.05;4 L2 {4 y$ h5 I( }9 W7 `
    
24. net.trainParam.epochs=50000;: {; `/ h- k4 }: h' b3 J  v  V, r
    
25. net.trainParam.goal=0.65*10^(-5);
    2 E9 Z( H2 z) p/ s6 [
26. net=train(net,pn,tn);
    \" r8 k5 f( @* V8 Z; C) n1 m/ ^) Z- W
27. 9 V# e; x+ X. _, g+ m6 P
    
28. %利用原数据对BP网络仿真& \' S\" V! r1 g! P3 ^# P
    
29. an=sim(net,pn);6 h8 ?$ b- p: z
    
30. a=mapminmax('reverse',an,ps2);
    8 k( i/ [# L0 n& v& z5 t+ q5 _
31. 
    % F+ d9 i7 S2 Q- n. {; G
32. %仿真结果与原数据对比测试
    7 H: X1 }! ]' H* t
33. x=1990:2009;
    9 k: Z1 m: y' P: ]& F& r
34. newk=a(1,:);
    : z$ W5 J- {( E+ Q
35. newh=a(2,:);
    # [: x* k$ n\" L9 t3 Y  c, L5 s$ ?  z
36. figure(2);7 h1 O+ D( n\" X( ^; @3 @
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    \" B1 j4 S! Z: i% S2 f
38. legend('网络输出客运量','实际客运量');
    * u5 W' q/ V. b1 ^, [2 {; S
39. xlabel('年份');ylabel('货运量、万人');: z6 m7 S% g/ d0 ?
    
40. title('运用工具箱客运量学习和测试对比图');2 B\" t9 d% D: @9 Z# j! Q: `
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    \" L  n/ u2 [# E8 S, E
42. legend('网络输出货运量','实际货运量');* d: {& B3 S& y& ]
    
43. xlabel('年份');ylabel('货运量、万吨');
    8 P. a7 `( E& g& J
44. title('运用工具箱货运量学习和测试对比图');
    # q) c3 E# c7 v  Y* x
45. 
    ' i8 \2 n' e* I\" A% {9 X, ^- Q& s
46. %新数据仿真7 E  U0 p. @6 o2 @
    
47. pnew=[73.39,75.55
    1 C. Y3 A. P5 c% i. p9 C# g& A
48.      3.9635,4.0975# d, l! m7 F! j: W5 O
    
49.      0.9880,1.0268];, u& k! P: W: P
    
50. pnewn=mapminmax('apply',pnew,ps1);. B3 m0 G; i\" R
    
51. anewn=sim(net,pnewn);+ P1 l# G. c7 _
    
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

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  ], }) Q) k, }( d2 ], i
   
2. net=newff(p0,t0,6,{'tansig'},'traingd');
   # j5 A. }$ R3 M- i' G8 C; m0 g
3. net.trainParam.show=1000;
   6 z' A8 T9 h' F' C. ]- l0 F
4. net.trainParam.Lr=0.05;& k$ H' I, b$ `* L9 o; H8 ~
   
5. net.trainParam.epochs=2000;; z! O* k8 n' J\" T1 p8 I% \
   
6. net.trainParam.goal=0.65*10^(-4);
   $ n) z5 [5 u7 ]7 y( T/ s
7. net=train(net,pn,tn);

复制代码

我只改了这里面的，
/ t* F( _! h* I, Y* K1 Q6 b一、改成单隐含层的，6个节点
. ^! W4 Q! ~3 s3 R$ K) _二、训练函数改成梯度下降BP算法 traingd
三、迭代次数改成2000
+ j% q$ Z+ p0 g% {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。
* @& Y  R, k, `6 S$ V( `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