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！！
1 g+ w& b6 y7 t' |: f
, L/ C$ C" j( T3 G' {; D! ^. a%原始数据输入
clc
) [8 q2 K0 _0 y% 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,...
# Z& T$ }( `, z1 ~    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,...
' [; g  L0 M6 [" b9 u, l6 e# b. \    2.5,2.6,2.7,2.85,2.95,3.10];
: t+ k- ]3 E+ X8 A0 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,...
    0.56,0.59,0.59,0.67,0.69,0.79];
4 `+ v) j, {! r: Kglkyl=[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];
8 d. e. S/ V; K/ K; qp=[sqrs;sqjdcs;sqglmj];
- B7 N( Q9 n+ I! {t=[glkyl;glhyl];
% u0 ?2 s% c: C+ w
%数据归一化
' q* R- N9 o( N1 w0 o[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
dx=[-1,1;-1,1;-1,1];

%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
. o* E% z9 A' P7 Nnet.trainParam.show=1000;
net.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
% _, x* a8 W+ S' U% F* F7 bnet.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);

( D4 f6 d/ Z, U) @. s) B# r; g9 ?%利用原数据对BP网络仿真
an=sim(net,pn);
+ J" B4 b  Z; I# Q9 b6 n+ v+ _a=mapminmax('reverse',an,ps2);

& i: q0 j8 u- z, ~, p5 _8 o%仿真结果与原数据对比测试
x=1990:2009;
- G8 \% h6 n! K- F  B' k  Z# }, onewk=a(1,:);
# E. J. [; C0 E% Xnewh=a(2,:);
. k0 w4 [. W, }1 R* {$ ^. ~! Pfigure(2);
0 r; J% Y8 I3 W# {2 h% Nsubplot(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');
xlabel('年份');ylabel('货运量、万吨');
; C, M( y9 L. G7 Stitle('运用工具箱货运量学习和测试对比图');
. l2 r* Z. M7 B! T5 {
%新数据仿真
pnew=[73.39,75.55
/ J; w+ t+ _# n" Y" `+ _5 {    3.9635,4.0975
    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
+ f; @5 P- `7 E; q3 lanewn=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 ...

0 J$ l( r# E! I' z* k9 U因为版本更新，Matlab中的Newff命令用法有所改变，原命令为：
( j3 ?% \0 i2 ^$ rnet=newff(dx,[3,7,2]{'tansig','tansig','purelin'},'traingdx');
7 S- _) _; i+ {4 E2 a7 o* A我修改为：
+ N0 ]. O9 v0 p. |* _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：
, k1 d7 X. t, D3 Z" s3 o[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
# L: g9 f1 c' ?5 Sa=postmnmx(an,mint,maxt);
6 Y* q5 Z1 m, w3 }% p- T/ b+ c# hpnewn=tramnmx(pnew,minp,maxp);
anew=postmnmx(anewn,mint,maxt)
. U4 v6 O7 B" w% X. j/ u7 Z我修改为：
[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
a=mapminmax('reverse',an,ps2);
pnewn=mapminmax('apply',pnew,ps1);
( G2 i7 Y! z7 j( d9 h9 N$ Fanew=mapminmax('reverse',anewn,ps2)

原程序为：
% e! N2 n' L8 f/ z%原始数据输入
clc
3 q, Z, v3 D0 \! I, fsqrs=[20.55,22.44,25.73,27.13,29.45,30.1,30.96,34.06,36.42,38.09,39.13,39.99,...
! x/ U7 ], |4 y& n& v- d7 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! O- r" Z$ ~$ l6 `    2.5,2.6,2.7,2.85,2.95,3.10];
% n, H1 s: c( F* a. Rsqglmj=[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];
( V% T5 P6 J' Q8 F3 Z, qglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
) @) Y1 [# p, h+ L/ C$ P# {6 G    22598,25107,33442,36836,40548,42927,43462];
: a0 b$ f2 O" O( W, Z: Nglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    13320,16762,18673,20724,20803,21804];
2 P8 g5 t  f  h9 l2 ~1 O- c7 Bp=[sqrs;sqjdcs;sqglmj];
5 i3 D, P! j! B% F! Dt=[glkyl;glhyl];
- ^: u! {/ I5 q) o
: y$ r0 f; a% k. x3 l%数据归一化
& h( r. G' h, u& |. }[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
dx=[-1,1;-1,1;-1,1];
1 q; o% d/ }. [/ ~
%BP网络训练
net=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
2 B7 d2 H4 O5 K  J/ @. v7 onet.trainParam.show=1000;
" h; t! ]2 f1 @7 Tnet.trainParam.Lr=0.05;
# V* }3 w9 F  J8 T. Anet.trainParam.epochs=50000;
3 K, g' v: |; O5 [" g9 U$ x0 F9 Qnet.trainParam.goal=0.65*10^(-3);
) k! a/ [  b, W9 R. |- rnet=train(net,pn,tn);

%利用原数据对BP网络仿真
an=sim(net,pn);
a=postmnmx(an,mint,maxt);
0 M& ^! y2 \, }4 W' z# a: p; B
%仿真结果与原数据对比测试
% ^- j$ ?" V! x# m/ Bx=1990:2009;
2 |$ a) r$ ^7 {# O; Rnewk=a(1,;
" ^6 j0 l9 O! }7 c9 v. g2 O2 g+ Dnewh=a(2,;
figure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
% z/ Z0 s) ]# k# [8 F" A6 glegend('网络输出客运量','实际客运量');
7 l9 ]- D& N2 ~" M+ E. txlabel('年份');ylabel('货运量、万人');
4 T( g9 t; I; V+ ^9 J' x# Atitle('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--+');
# w- \+ H: `& k% [: b. slegend('网络输出货运量','实际货运量');
+ d: z& E- y1 K0 M7 M/ Nxlabel('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');

%新数据仿真
/ P0 s& }( e7 t2 j* E5 Cpnew=[73.39,75.55
! s: P+ |# C) B  g! R' Z% G: H    3.9635,4.0975
    0.9880,1.0268];
% u1 y0 L# [3 l3 S. npnewn=tramnmx(pnew,minp,maxp);
anewn=sim(net,pnewn);
& V& V9 P. w8 ?# }anew=postmnmx(anewn,mint,maxt)
1 c9 D/ D% e+ h" Q
( A3 v  j) o" H! A6 g/ c- \修改后程序为：
%原始数据输入
clc
7 |2 B7 @8 b$ J8 |% B0 a& 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,...
& n) D% Z( P% M3 O2 y+ o& ^- y    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
/ @- w# Y+ ]' m8 I3 gsqjdcs=[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];
* }$ r' P8 U* p6 v+ Isqglmj=[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' a# D8 b- w% O2 i+ e0 |    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];
glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
$ y7 c8 ^9 D, o    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
! T2 u+ f. B9 L: z9 ]t=[glkyl;glhyl];
5 i9 R! z3 {4 W& F$ B8 {8 p
%数据归一化
[pn,ps1]=mapminmax(p);
% v8 Z( c+ `0 k: e[tn,ps2]=mapminmax(t);
+ K, b& b( N! V  J9 q
%BP网络训练
# p$ B& k1 u: F/ I0 Vnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
" c" ]/ k+ P4 L+ b1 Znet.trainParam.show=1000;
6 F4 O" |% q& E" ~* I; H' l" x% Onet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
9 [8 g4 a  o4 Z, Tnet.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);
# k! E' v3 |9 c% v7 q0 ^4 wa=mapminmax('reverse',an,ps2);
! [( Z* d& \7 {/ y( E7 Y) v' j* |
%仿真结果与原数据对比测试
0 u  U8 }# w8 K0 s/ Ux=1990:2009;
: Q) }5 q) [8 X. f, [% inewk=a(1,;
  j' p8 M8 Q  _1 n$ [& y5 wnewh=a(2,;
: z9 Q8 ~' }/ A  }- T/ Y" t2 ^figure(2);
8 D6 n+ h; F/ m3 jsubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
8 p# C3 {( L; }, xlegend('网络输出客运量','实际客运量');
4 c; b8 R4 d( H: @1 Z- {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');
3 k! o: i  a) {* d- W' n9 G* nsubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
, t! j: [. z  U5 {legend('网络输出货运量','实际货运量');
+ Y7 A  E5 [1 s* z4 V. }5 ]xlabel('年份');ylabel('货运量、万吨');
2 W6 }( L% ~" m) ?- etitle('运用工具箱货运量学习和测试对比图');
" p5 q4 k$ C, j1 T
%新数据仿真
: g' A+ d# y0 F$ a* p1 epnew=[73.39,75.55
3 a0 @" I1 W3 x6 r1 o( q. Y% r    3.9635,4.0975
    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
2 L6 n. W) d+ _& u9 D- q（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. clc6 p1 I. U\" N% o
   
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,...
   $ V$ i- k. L6 ~: ?; E
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];* j! j! ~% i$ k7 I
   
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,...# m: [7 E2 ], A$ N0 x( p# H+ |1 o, }
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   0 S& y, g7 v; ?0 r/ u) C  ]\" r
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,...3 j* F, P; X8 S8 R
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   ; v3 \0 M, L7 |' c& q+ [
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...7 C' x1 Q, U7 Q) P& Q: e! v& }$ N
   
9.      22598,25107,33442,36836,40548,42927,43462];
   * \; g  a% C\" B% `1 u% o; ?
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    * ^, f* I$ d  g. A
11.      13320,16762,18673,20724,20803,21804];2 f\" S. H  C1 ], E4 S
    
12. p=[sqrs;sqjdcs;sqglmj];( H6 n, u' z* \\" M
    
13. t=[glkyl;glhyl];
    ' f5 V! _* i0 N9 K9 Y\" s) F
14. %数据归一化. j; W7 i) L\" A& }: }
    
15. [pn,ps1]=mapminmax(p);1 S/ i/ h1 v8 N# ~
    
16. [tn,ps2]=mapminmax(t);
    ) N+ h4 R/ W. j1 q
17. %dx=[-1,1;-1,1;-1,1];
    5 h. j3 G: U' R/ K- A* x5 E
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]5 h9 C0 W9 N1 q4 |& H, `  H
    
19. - a* j1 F8 ~# W4 v6 r9 a
    
20. %BP网络训练) o' G7 a, S' \3 e. A
    
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]
    4 X\" _$ b* o! m' i
22. net.trainParam.show=1000;! O; G% ?4 a1 |5 W
    
23. net.trainParam.Lr=0.05;
    $ {$ f9 x5 w\" C  i5 `1 Y' h8 R0 R
24. net.trainParam.epochs=50000;
    1 W0 N' N% d% h% M3 d2 X
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]
    \" j$ o: Q$ H3 S: t
26. net=train(net,pn,tn);( Y' ]9 c( ~% X7 i# w* u
    
27. ! u4 b; x/ h9 }+ n) W- q$ M
    
28. %利用原数据对BP网络仿真
      w1 u7 L; w) D) a8 v$ m& q
29. an=sim(net,pn);
      q5 F; A6 }! m
30. a=mapminmax('reverse',an,ps2);
    0 j' {/ E- c2 D\" h- C, ~3 z5 `' V
31. 6 X& D4 R$ S% I- a
    
32. %仿真结果与原数据对比测试 ) w; m9 E- O4 a5 U
    
33. x=1990:2009;
    . H8 Y3 z; L! c. z% c- Z
34. newk=a(1,:);: u5 s. X) P0 [! O! L- H
    
35. newh=a(2,:);
    7 v* g7 ~2 c4 h6 U  k% N6 E) Y
36. figure(2);
    ! h0 M- {0 ~& N* z
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');- W/ }. ~/ e. Y) @8 r* A' F
    
38. legend('网络输出客运量','实际客运量');- S5 d( Z4 N' r9 a, \: l4 N3 M
    
39. xlabel('年份');ylabel('货运量、万人');
    * ]$ }1 C. v+ T+ o4 ~2 K4 T
40. title('运用工具箱客运量学习和测试对比图');
    5 T$ A. f. u6 `
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    / Y+ e* _$ Y0 Y  F
42. legend('网络输出货运量','实际货运量');
    4 n+ N7 w\" R# H; k0 n8 T% d
43. xlabel('年份');ylabel('货运量、万吨');
    6 Z) _: \3 @2 L' K' {' r
44. title('运用工具箱货运量学习和测试对比图');
    0 I* F( V' N4 t
45. 
    / w+ [% \/ \3 F' [# p* U
46. %新数据仿真3 _- ^+ @2 ]9 m( N/ ]
    
47. pnew=[73.39,75.552 T5 _0 K3 r3 R$ e+ ?# f
    
48.      3.9635,4.0975
    9 y\" ~# b7 D. f& F  G3 ]# x
49.      0.9880,1.0268];
    . A# S9 j8 P& R6 {: K
50. pnewn=mapminmax('apply',pnew,ps1);
    6 T% c4 p& F$ W9 N/ i! p3 A
51. anewn=sim(net,pnewn);# n. b- U; o( ~% }( K$ n0 N- a8 o
    
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
   6 Z% y' F' {4 w' ^5 x! Y  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,...5 P0 [4 K8 t. Q5 e/ [/ @6 B\" y
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];2 T$ Q* w1 a\" T8 }1 V1 t
   
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,...) Z- b! w: [\" c\" ~) D\" s; {* r
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];* L  f\" H0 R5 W2 C2 }0 a
   
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,...
   % i2 O% V' y  s) d\" l$ a9 x, H
7.     0.56,0.59,0.59,0.67,0.69,0.79];, v# E3 u. s% f; ?/ n3 A3 N
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   5 I5 e' M5 s! T$ l# `$ d& t& b
9.      22598,25107,33442,36836,40548,42927,43462];4 ?5 e' b# a  d8 V8 {
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,..., ?8 ?# h7 v( n5 T( G/ _
    
11.      13320,16762,18673,20724,20803,21804];0 s6 d+ M- P, J% h6 W
    
12. p=[sqrs;sqjdcs;sqglmj];& c7 l; M' X  [
    
13. t=[glkyl;glhyl];
    : O0 t4 A; @9 p0 u7 [/ }/ J
14. %数据归一化
    2 Z5 {1 z! O( r0 f: j! H
15. [pn,ps1]=mapminmax(p);( q3 U, I, P5 D4 M
    
16. [tn,ps2]=mapminmax(t);
    . g5 |' U$ J- F1 w\" O
17. %dx=[-1,1;-1,1;-1,1];, l, r6 o6 ?0 k' j
    
18. p0=minmax(pn);t0=minmax(tn);# n- [0 Q& ~8 w* ?
    
19. 
    2 X+ i, S3 Y; l; Z, G# B% z5 ?
20. %BP网络训练) A4 r+ _7 m: p: x$ W+ M, y- m. @
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
    ; X0 D7 `# k  f3 i# i3 B4 O\" b' K
22. net.trainParam.show=1000;: k- n5 y  P, D) R* [
    
23. net.trainParam.Lr=0.05;
    1 `5 a1 r\" D) \& x5 b
24. net.trainParam.epochs=50000;% m, ?5 U! x\" L* F1 [7 x9 \
    
25. net.trainParam.goal=0.65*10^(-5);5 c' {\" y, x5 b& h7 R# p
    
26. net=train(net,pn,tn);
    3 `# [+ N, _1 _9 H
27. ) [  B0 p  @, b\" O4 j+ ~
    
28. %利用原数据对BP网络仿真
    ( v$ }/ T3 |7 B+ }3 m4 o# E
29. an=sim(net,pn);\" B7 Q/ a' Y  m\" A- z- Y% d
    
30. a=mapminmax('reverse',an,ps2);6 V- B2 y( A! n/ z6 k
    
31. 7 N/ y; `8 f4 d' h\" ]
    
32. %仿真结果与原数据对比测试 , I: L\" ]& E2 ]* P\" b) L5 M
    
33. x=1990:2009;
    3 v4 Q- R# B8 X  `; y+ d
34. newk=a(1,:);, F* @\" Z1 q$ x4 C7 D# i7 q# A
    
35. newh=a(2,:);
    4 S( c( t4 e1 L! b
36. figure(2);' j0 T% O* ]; T
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    % n- b, H! }& G: ]$ b# l
38. legend('网络输出客运量','实际客运量');0 k4 h( e! m9 e$ k( W' h6 s0 I7 W& f
    
39. xlabel('年份');ylabel('货运量、万人');
    5 Y3 [: S; d5 Q/ m
40. title('运用工具箱客运量学习和测试对比图');
    / J- v/ ~, f7 D7 e# V
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    + B9 @$ w\" F& `: Z/ j% O\" W4 ?
42. legend('网络输出货运量','实际货运量');
    \" [3 p8 B5 @' M9 A. [# |
43. xlabel('年份');ylabel('货运量、万吨');
    2 `\" M/ f& {! t) R, c. ~! ?% W
44. title('运用工具箱货运量学习和测试对比图');( q3 `1 h) t8 R! t
    
45. 
    5 G& s& \/ r' }& W1 S& v
46. %新数据仿真
    $ _& \' D# W# \' o2 d% K- s
47. pnew=[73.39,75.55
    ! q! u8 c  B: |
48.      3.9635,4.0975
    ! c# D: U  z2 j& o+ E
49.      0.9880,1.0268];
    / E* s$ a& ?. y6 u! |( c( T
50. pnewn=mapminmax('apply',pnew,ps1);
    5 f- P7 W  a5 K) z
51. anewn=sim(net,pnewn);
    0 g5 `. n; l* r4 H! h5 s. k/ N
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
   + {; }* P' U3 ^- ~$ P\" ~+ [& _
2. net=newff(p0,t0,6,{'tansig'},'traingd'); % b# s$ x  C2 X9 I1 k$ e3 y( l
   
3. net.trainParam.show=1000;, X7 k+ ~( T% v: m
   
4. net.trainParam.Lr=0.05;# G) A5 v/ Y# }$ ?4 F$ {# T\" h% @
   
5. net.trainParam.epochs=2000;
   * v7 ?\" y$ `  c, j+ S
6. net.trainParam.goal=0.65*10^(-4);2 O- W6 r0 P& o' |
   
7. net=train(net,pn,tn);

复制代码

我只改了这里面的，
! L/ z  Q4 @- D3 e一、改成单隐含层的，6个节点
/ S# g4 V: ^8 C二、训练函数改成梯度下降BP算法 traingd
/ t' j" D# m0 d" V/ ]三、迭代次数改成2000
" i; G, n* E6 N上面的参数是自己试的，我也不知道为什么。
! |4 z! `7 F! @4 W  P' ~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