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！！
7 W+ u- Z, K8 A" v- d: W, t5 {
%原始数据输入
( a* O1 t0 K8 l( V* [% ~. L9 ^8 M; }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];
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,...
/ n6 f) v3 z0 Q5 B+ g! Z/ @9 r    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,...
9 e( T9 {3 C* C# O3 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,...
    22598,25107,33442,36836,40548,42927,43462];
glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
0 P4 S( B! Y1 |    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];
+ U2 e1 E  J1 ~+ f. n
6 Q2 X5 J9 u; X9 \0 N9 s1 o$ g, D%数据归一化
* |5 u, p# D4 A+ q[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
7 N  E! o7 f- P& K% X2 Udx=[-1,1;-1,1;-1,1];
0 I- m. p( {/ K  ]5 f. J) T: G
%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
' A; ]( Z: O6 S# y: j& X& qnet.trainParam.show=1000;
2 C4 x3 w, c3 W7 A/ T9 o+ Q2 ^( Enet.trainParam.Lr=0.05;
4 G) m  X. T1 [' `! M! {+ Ynet.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
, l- G; u4 E- z/ [an=sim(net,pn);
& c% ]  U) I8 b/ _/ \7 Ra=mapminmax('reverse',an,ps2);

%仿真结果与原数据对比测试
x=1990:2009;
' l3 ~2 ~' ?' _' F. O! I2 g5 Q1 snewk=a(1,:);
- O6 R5 Y; c* p: o0 anewh=a(2,:);
: Z5 D8 T& C& s. b( ^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');
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--+');
. s% w- [0 K; q  o& k2 Wlegend('网络输出货运量','实际货运量');
$ _( o. u" @) ?+ L+ r% Axlabel('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');
& f$ k3 W4 q( A/ `5 P
%新数据仿真
1 Y4 A: w5 }5 ^9 T) Y2 j! wpnew=[73.39,75.55
) ]& T4 V! X$ o    3.9635,4.0975
    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
# ^  x% t8 S1 h3 J# v' ^anew=mapminmax('reverse',anewn,ps2)

+ p0 `# h) e0 D  R2 y( U* w

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
: s. @& Y; ^8 v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 ...

' H$ ^- J. S& Z/ x- A+ B$ x  V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：
% n6 c6 B; P, Inet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
, {4 {5 P5 E2 ], V' r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);
" M. q- ]4 V# d  Ca=postmnmx(an,mint,maxt);
" y" d; C; u4 _& |* W1 Dpnewn=tramnmx(pnew,minp,maxp);
+ A! b' ~* C: m) e0 X0 s5 vanew=postmnmx(anewn,mint,maxt)
- @7 }; k6 R% N& s我修改为：
6 R" L) w7 y/ l4 N4 g[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
8 H2 r) \- ?9 [( G% u. oa=mapminmax('reverse',an,ps2);
pnewn=mapminmax('apply',pnew,ps1);
% m9 c$ Z9 [* ~* w6 kanew=mapminmax('reverse',anewn,ps2)

原程序为：
0 L# h1 Q. x9 D%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];
2 [- ]# S8 }5 E7 rsqjdcs=[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,...
8 i+ E/ P0 w- P; A4 c/ F% Z% z    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];
1 e" m/ k9 |5 a3 k9 t) Wglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
0 T4 ^2 D+ N5 N  u6 S    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];
) t6 W/ f( I3 O4 m0 `
%数据归一化
[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
6 y: b* I) E" j) bdx=[-1,1;-1,1;-1,1];
; t* A2 V' A2 i$ P  R6 i, G' C1 @
%BP网络训练
net=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
7 O9 A0 t$ Y. s, o* x7 vnet.trainParam.show=1000;
$ }# i, T- @1 Vnet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
$ g8 j3 E" V% h" X% O' {net.trainParam.goal=0.65*10^(-3);
5 W% f0 c" V: ^: M7 Jnet=train(net,pn,tn);

3 P6 s  R# q. x5 O%利用原数据对BP网络仿真
5 s% _) t+ o, D7 k4 l5 wan=sim(net,pn);
a=postmnmx(an,mint,maxt);

4 ?2 E6 q( S+ B! S+ E7 `. N! n%仿真结果与原数据对比测试
8 y) V) O% W" I" K  yx=1990:2009;
  N% C8 C* \7 f  ?- Tnewk=a(1,;
newh=a(2,;
3 y1 |% M, u6 Mfigure(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');
' D5 u. g, r3 j* {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');
  F; t4 s: o9 f; Etitle('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
* @1 m9 i* Y; ^1 m    0.9880,1.0268];
pnewn=tramnmx(pnew,minp,maxp);
/ N* f$ y9 U, T7 o, hanewn=sim(net,pnewn);
0 l* e* ?) ]: [; s9 _anew=postmnmx(anewn,mint,maxt)
! R& ]/ A0 s' H$ N2 o% F
修改后程序为：
! R& w) b6 Z5 l) ~, f- K; t%原始数据输入
' v' \# q7 ~. G+ y8 tclc
7 ^( I3 c3 h/ {1 ]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,...
$ e+ ?; s& A5 K2 }1 Y    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
: R; p$ v, \& V4 I2 msqjdcs=[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,...
, P6 k3 [7 v: w% H* \    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,...
& R: I" m% f- M    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];
; k. I0 H0 C- F+ {- i( Pp=[sqrs;sqjdcs;sqglmj];
+ ]/ B) {5 {, Nt=[glkyl;glhyl];
( }* ^/ U+ A: _$ s0 p
2 M8 U' {0 I  m! z5 _9 ^2 a* L) W%数据归一化
[pn,ps1]=mapminmax(p);
; d1 X' N7 y' c- K. I( P' f[tn,ps2]=mapminmax(t);
0 M5 w! u5 o! ^# \/ [
%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
8 }9 O1 R; [+ l( @net.trainParam.Lr=0.05;
. c( H9 }& k! p5 {net.trainParam.epochs=50000;
( ~  `, }1 ]( x: S/ i" _) |net.trainParam.goal=0.65*10^(-3);
3 G: S' c4 O& S% u7 Tnet=train(net,pn,tn);

3 p  s9 R8 b5 P$ I1 m9 `  o%利用原数据对BP网络仿真
an=sim(net,pn);
  E$ I* x& X1 `4 i  y1 Z! w& ma=mapminmax('reverse',an,ps2);
2 w. D3 D# ?5 q, t9 z6 f/ F
4 h3 g, i- T$ i%仿真结果与原数据对比测试
' L- Y0 R% ?- X6 D, }3 B4 ?9 S# Ox=1990:2009;
) e- ]. }) n' A4 u2 Hnewk=a(1,;
newh=a(2,;
% Y* W; U0 l2 B, m" ?" Yfigure(2);
) s! Y7 Y# ?2 l$ \2 {+ p2 _  k& Q( b; msubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
legend('网络输出客运量','实际客运量');
) Z- C* [% h& m' u$ _' I8 g* hxlabel('年份');ylabel('货运量、万人');
2 @8 o4 `% Z$ ktitle('运用工具箱客运量学习和测试对比图');
$ t% x+ I/ G6 M* N, G( Csubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
7 F8 O. P0 D4 o. z, f2 Glegend('网络输出货运量','实际货运量');
. v  s! J. f: ?% f( h" gxlabel('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');
, ?8 X% p; d3 E
+ m5 S+ n- T: {0 ]( p' T%新数据仿真
pnew=[73.39,75.55
1 C2 ^" L/ P( N    3.9635,4.0975
8 G- o* c. B$ M6 X    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
4 v( y" f# m% Z1 \( A# h; y（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- a- j+ w' r, Z+ @$ i# u1 V5 y' n
   
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,...\" i, c\" S* O2 \
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   2 n3 f. e* f: g
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,...
     a) U% K6 E' N
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   : R. G1 l0 v: ~3 R7 s
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 T2 L5 {3 S$ f
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   . @  U8 }3 p- ?' f( ]
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...9 n! G# w/ t\" _
   
9.      22598,25107,33442,36836,40548,42927,43462];
   6 I& D. g1 I! I7 q0 f9 f
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    * Q6 h! f1 |* c* \+ O% b  C; _
11.      13320,16762,18673,20724,20803,21804];
    7 B- ^1 j- s\" q3 K, T\" w0 S& k
12. p=[sqrs;sqjdcs;sqglmj];- o9 x5 n( |# G4 i4 W  k/ x6 J! W2 m& T
    
13. t=[glkyl;glhyl];
    # \9 h7 T  Y3 ^& h/ R
14. %数据归一化
    ' e\" G6 Z: U$ k; D0 O1 J
15. [pn,ps1]=mapminmax(p);
    # b) [. {+ w% W8 a9 ~' P6 i
16. [tn,ps2]=mapminmax(t);
    ) _3 q) a/ F) H) n0 h8 F
17. %dx=[-1,1;-1,1;-1,1];
    2 K3 S! |% v( z& E
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]- Y& ^% [0 S% J+ Z3 J  S
    
19. : K; u# G, j3 d! A( e
    
20. %BP网络训练
    ' _* I, X* V& D9 y, n\" d
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]
    + o! R  E4 \- X5 S- v
22. net.trainParam.show=1000;8 w0 [5 y0 h! n+ U  c$ Y$ |/ Z
    
23. net.trainParam.Lr=0.05;
    & |! W( ~$ r3 p7 l! O4 f8 f) G* `% Q
24. net.trainParam.epochs=50000;
    # i2 i+ h$ L$ V' }; L# X* {
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]
    4 j2 Y. b5 c: X3 e  Q8 [6 o$ \
26. net=train(net,pn,tn);
    5 @9 n6 l: [' [# O( h/ w\" O
27. 
    7 F. C, \- |. s
28. %利用原数据对BP网络仿真: ^& T  r\" j6 I\" l) m1 u
    
29. an=sim(net,pn);
    $ v5 _& G; Y6 V4 n5 j. w8 R
30. a=mapminmax('reverse',an,ps2);+ K) ^; l& S( Q
    
31. 
    / e- V: _# s8 _% f/ G1 q
32. %仿真结果与原数据对比测试
    / X0 a; I: E0 U* i0 i- e% i: |
33. x=1990:2009;
    - ], n5 s+ H\" `; X  B' Q) v2 _
34. newk=a(1,:);
    \" b. A! h; n  r( e
35. newh=a(2,:);
    / e/ F4 B- p) c6 T
36. figure(2);
    * L' I9 M/ x' |$ A# O9 j: R+ v% G
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    2 e+ P\" s; s) F- `
38. legend('网络输出客运量','实际客运量');) K+ R# m0 {  W5 ]* H+ c
    
39. xlabel('年份');ylabel('货运量、万人');& D( k& R3 y2 ~2 P\" c  ~
    
40. title('运用工具箱客运量学习和测试对比图');
    7 I( ^2 w! L$ X& T
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');, ]7 ^$ \% N5 e
    
42. legend('网络输出货运量','实际货运量');
    & @: E+ v) I0 F3 f; D# J
43. xlabel('年份');ylabel('货运量、万吨');
    5 L5 t# B3 v) q- f2 a* V! H- v
44. title('运用工具箱货运量学习和测试对比图');: |! {; K\" b% E4 T
    
45. 8 F$ s8 ^, U# O7 W; v: @
    
46. %新数据仿真
    , B; y) T1 S% q1 E
47. pnew=[73.39,75.55% ~, d- D\" \; h' W6 P+ A7 L
    
48.      3.9635,4.0975
    # I- y6 N) u. P; g6 ^( m+ ?
49.      0.9880,1.0268];3 p/ i7 G* H, q2 f. d! B9 Y( m
    
50. pnewn=mapminmax('apply',pnew,ps1);
    + d. N0 @8 C$ [2 F& G* Z/ Q6 t
51. anewn=sim(net,pnewn);7 G6 O& ^4 ]1 u' r* g+ `& K9 z1 P8 [4 ?
    
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, C% y# q5 n4 I1 Y+ l
   
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,...1 O: Z! O' J/ A
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];- r/ w1 B\" ~  u  S$ z7 p5 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 ~* w6 Q7 v3 e4 t4 ~7 R& v
5.     2.5,2.6,2.7,2.85,2.95,3.10];
     w; v) W, W6 I7 d
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,...# p+ Y) E) {0 c4 ^5 I4 |
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   & E: _- d) \! S* V! C9 ]) @2 R
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,.../ t2 ]  M4 x$ y( o5 z* \
   
9.      22598,25107,33442,36836,40548,42927,43462];# u. S0 J2 v5 e
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    8 D- f4 b  o\" V, E
11.      13320,16762,18673,20724,20803,21804];
    % }( |. n5 G$ }. u
12. p=[sqrs;sqjdcs;sqglmj];
    ' j  N% T\" q* k- ~* S2 j, h
13. t=[glkyl;glhyl];+ m( _& K8 N' x6 B1 S. t: v: N
    
14. %数据归一化8 G1 _6 q/ j! n\" W0 s. f: a
    
15. [pn,ps1]=mapminmax(p);4 D1 T# B( }2 ~# {# E
    
16. [tn,ps2]=mapminmax(t);1 `: T\" t/ M) k* B
    
17. %dx=[-1,1;-1,1;-1,1];1 A; s5 a\" C+ V- S
    
18. p0=minmax(pn);t0=minmax(tn);+ ], f2 v& q6 T5 F2 i- g7 X4 x
    
19. 
    $ D  ?6 P8 Q' `2 \% T. z3 p; p
20. %BP网络训练( S/ v1 O# n( d2 {* }* g6 C
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
    % B* k! |1 ^# A6 F& L
22. net.trainParam.show=1000;0 _4 H- z, d( l. Q' T* b; J
    
23. net.trainParam.Lr=0.05;
    3 o, V: a% p; F% `2 I. a
24. net.trainParam.epochs=50000;
    ! c* V  H  T% n. a, c* s
25. net.trainParam.goal=0.65*10^(-5);+ x6 S8 N+ G. z5 n
    
26. net=train(net,pn,tn);9 }7 o4 ?  d; d  j( i9 X  F7 N1 M, m' X
    
27. 
    / g$ t1 O1 R& A! U+ s3 h4 H
28. %利用原数据对BP网络仿真, c* [& V% o& k7 Z/ _7 q
    
29. an=sim(net,pn);, h; _8 N8 u0 J3 u5 C
    
30. a=mapminmax('reverse',an,ps2);
    9 s! L* u2 {; ^' g; v( }
31. 4 }8 F\" y\" S! U  x
    
32. %仿真结果与原数据对比测试 % n( r- C. D8 I: |
    
33. x=1990:2009;
    , f7 I0 |6 M; z( a
34. newk=a(1,:);
    . d. g8 ~5 J* c4 R6 \
35. newh=a(2,:);! ?' J3 F6 M+ j0 y( a
    
36. figure(2);
    : P4 j4 T/ o  K5 Z( `
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    \" x; p7 p% F  U( L' }0 Y$ f! G, D
38. legend('网络输出客运量','实际客运量');
    5 C+ k7 ]  M/ u
39. xlabel('年份');ylabel('货运量、万人');
    5 F( G& `5 x9 ?. _0 W
40. title('运用工具箱客运量学习和测试对比图');
    \" L4 W! ~& c' }8 i
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');# k5 X. Q' Z* [' J; K
    
42. legend('网络输出货运量','实际货运量');
    4 _\" a- K( F$ N* Q
43. xlabel('年份');ylabel('货运量、万吨');
    # a* Y2 s# o9 d
44. title('运用工具箱货运量学习和测试对比图');/ R/ a( D: W% ~* X# j( I
    
45. 
    5 [/ Q7 Z: X1 Z8 L) }) {) b' t+ J
46. %新数据仿真
    4 p* F. T- T# A1 c1 q2 c% J
47. pnew=[73.39,75.55
    / `1 ^1 l; B) G
48.      3.9635,4.0975+ A$ F1 o& s( j! w% A
    
49.      0.9880,1.0268];2 J) \5 E' v. p5 x
    
50. pnewn=mapminmax('apply',pnew,ps1);% O* i& O1 T$ x/ N) q. v
    
51. anewn=sim(net,pnewn);! T% N! K! x, P3 T5 d* K, `
    
52. anew=mapminmax('reverse',anewn,ps2)

复制代码

T-Eric

且生 发表于 2014-7-21 18:22
5 `6 u& F+ O2 {. f0 ]' M* E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
   ! Y! _; H( m6 ?+ L& N
2. net=newff(p0,t0,6,{'tansig'},'traingd'); ) H7 {+ ]6 x3 p3 ^8 ]
   
3. net.trainParam.show=1000;7 i, j. ]( {5 p3 H3 F
   
4. net.trainParam.Lr=0.05;
   ! l\" `5 U7 N  i
5. net.trainParam.epochs=2000;
   ) @6 y9 q! y  |9 y& {
6. net.trainParam.goal=0.65*10^(-4);
   # ^! ^$ ]8 ]\" D# ^0 f7 K: p
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
& V  n- z& u6 N9 x5 c% ?( Z二、训练函数改成梯度下降BP算法 traingd
9 I5 ~7 E: Q( S- |; j三、迭代次数改成2000
) n0 u% Y( V/ q6 _9 a  `4 }上面的参数是自己试的，我也不知道为什么。
+ P7 x9 I+ [# h/ f- a9 \$ }6 L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。
% H$ l' m& C9 Y7 G; S2 X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