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！！

%原始数据输入
0 O; t+ f$ g/ A8 |9 j$ d7 F! i% bclc
' j& O* h  f% d$ A3 [3 ~" M* D/ 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,...
- Y1 O! |9 I9 D9 F1 e+ i  _. v    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
8 ]$ U5 ?7 _/ M. [# [3 ^6 v+ 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,...
    2.5,2.6,2.7,2.85,2.95,3.10];
2 V+ n% }& e. Z" i2 i( O& I! ~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];
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,...
( d  g' z3 i+ D* e7 X" P- p    13320,16762,18673,20724,20803,21804];
3 |) i( D6 q& X. N! }# S7 ^1 G2 Kp=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];
) m7 \# r' Z! p" H1 m0 r, }
%数据归一化
3 V$ b, z* M+ Q' p[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
. s* a7 G' _: o# f. f6 \dx=[-1,1;-1,1;-1,1];
8 o: v, n$ _8 j1 s: W  k7 X, `3 U
9 L1 n$ O$ ^8 R* L; d%BP网络训练
0 T- ?% P1 f. {4 jnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
( s* ~: X) g# c$ S+ B1 q5 f! snet.trainParam.Lr=0.05;
/ m0 j0 p& h/ ~- A  Tnet.trainParam.epochs=50000;
! W4 P& l% ?; D. snet.trainParam.goal=0.65*10^(-3);
. ]6 t+ V# q) m& enet=train(net,pn,tn);

6 w2 Z  L9 K0 f& X%利用原数据对BP网络仿真
an=sim(net,pn);
9 t. z- o  A6 v: h8 Q; p# Ja=mapminmax('reverse',an,ps2);

%仿真结果与原数据对比测试
5 _2 q; ?) \& u: v9 \x=1990:2009;
: H3 s, }7 G: F7 O2 C% [" x8 ^newk=a(1,:);
$ e5 x) n* H2 R& m* r& knewh=a(2,:);
figure(2);
. y' l' b; K  U6 }' Msubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
: |/ O8 g4 T7 C, i8 blegend('网络输出客运量','实际客运量');
xlabel('年份');ylabel('货运量、万人');
9 T8 O! D8 T( o8 D- ~) dtitle('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');
. X8 v! o; c# f. ]' }: M2 Y$ W( ntitle('运用工具箱货运量学习和测试对比图');

; @$ \; i; i3 q" G- q. P%新数据仿真
/ c: A) U1 k) g: epnew=[73.39,75.55
6 W& E/ d. r: l    3.9635,4.0975
    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
1 Z. }2 {4 ?0 r+ C/ }, k

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
* a) q) F0 [% E+ M% J# V& I' ]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 b" h# k  n! {因为版本更新，Matlab中的Newff命令用法有所改变，原命令为：
) f# P: Y% j7 |$ inet=newff(dx,[3,7,2]{'tansig','tansig','purelin'},'traingdx');
. O6 |8 }: }* I, \' Y8 a( v# e我修改为：
5 J% D4 V; K& J0 z! E/ a7 enet=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);
" B4 I6 C$ x9 R/ y* i' J8 z" E. Ka=postmnmx(an,mint,maxt);
pnewn=tramnmx(pnew,minp,maxp);
anew=postmnmx(anewn,mint,maxt)
4 V) L" n6 w5 U3 M& l我修改为：
8 C7 m$ v$ a  m. T: g[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
a=mapminmax('reverse',an,ps2);
# C% O6 O, W; F4 Lpnewn=mapminmax('apply',pnew,ps1);
1 l4 a% u8 D7 G: `' `% X9 X3 nanew=mapminmax('reverse',anewn,ps2)
& j: V0 C0 Q- _; ^" J, }/ S0 S6 S
6 r& n9 `1 T, p- Y3 Y原程序为：
%原始数据输入
clc
3 n7 O3 v5 ?: [5 s$ x! [, Asqrs=[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];
/ v3 N8 {- B8 D& `1 f- ~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,...
  ]7 Q5 e' g8 C* X; {    2.5,2.6,2.7,2.85,2.95,3.10];
1 E% C. I! G0 p* Y, c9 vsqglmj=[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,...
( \/ z5 q9 [. n. e  A    22598,25107,33442,36836,40548,42927,43462];
# u- C: ]% R% m' `glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
8 U: f1 v1 b1 h, ^    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

%数据归一化
! Z1 i  W! R9 G/ p' P* F; `, z3 Q[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
: A7 Q$ S  f9 h9 Odx=[-1,1;-1,1;-1,1];

' h/ p  @! A$ X4 z%BP网络训练
net=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
- u, q3 U! N- A; y$ Pnet.trainParam.show=1000;
net.trainParam.Lr=0.05;
net.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
an=sim(net,pn);
8 }4 S3 ]7 c6 T1 p# x9 k4 ha=postmnmx(an,mint,maxt);
5 A; c' m& S1 g3 j, B
%仿真结果与原数据对比测试
x=1990:2009;
& m1 ?, d; s9 Y% ^8 knewk=a(1,;
newh=a(2,;
7 [' G' o7 r8 dfigure(2);
7 p+ V7 B/ J, t& U; J  Zsubplot(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');
4 F* y& o) W0 |! w% i+ z4 u' btitle('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');
; s# O) E3 q2 Y6 O! v& Cxlabel('年份');ylabel('货运量、万吨');
, b8 M4 i; u- }1 I/ Y6 R, b6 r6 Stitle('运用工具箱货运量学习和测试对比图');

%新数据仿真
) M0 M+ C" N' i+ Gpnew=[73.39,75.55
4 r. V; U% N- j    3.9635,4.0975
/ y  m- l4 _) p; N    0.9880,1.0268];
pnewn=tramnmx(pnew,minp,maxp);
. j3 a- E) h* p. ~6 danewn=sim(net,pnewn);
. N+ e. U0 w9 x6 k7 ^anew=postmnmx(anewn,mint,maxt)
3 B, c" @/ q+ n
修改后程序为：
8 A# `* r0 t' U4 j4 I$ S+ X0 c" O%原始数据输入
0 s. a, P' r  T" N- |' yclc
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,...
2 V0 G& s6 E) ^1 L5 F    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
( e+ n9 P4 p  H. ^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,...
% E0 y; j4 p3 ?5 ^6 K    2.5,2.6,2.7,2.85,2.95,3.10];
# T$ w+ t9 ~3 nsqglmj=[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 e7 Q1 Z9 p% I0 s* \3 F6 g& [( J    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,...
9 O: P8 i; B+ H    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
6 H/ G- a* S, a- ?, \: ut=[glkyl;glhyl];

%数据归一化
2 I" [% I6 @: T[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
3 J' j: S4 s  T  f& c: l$ l0 {
%BP网络训练
! p% @- x, [3 Y3 ^- qnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
net.trainParam.Lr=0.05;
; ^; c+ `- u% W; A$ ^! Xnet.trainParam.epochs=50000;
1 z, R4 o1 ]) S9 q9 k6 [net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);

" G9 o+ u$ I  i! J3 N+ @4 W7 a%利用原数据对BP网络仿真
an=sim(net,pn);
a=mapminmax('reverse',an,ps2);

%仿真结果与原数据对比测试
x=1990:2009;
. l$ H* J3 r9 Qnewk=a(1,;
newh=a(2,;
9 p/ [: h1 p9 O7 yfigure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
9 w1 I: b7 ?1 n  Elegend('网络输出客运量','实际客运量');
xlabel('年份');ylabel('货运量、万人');
" V) D$ p7 [2 i; ytitle('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');
' p: X+ S. D0 b2 |- Yxlabel('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');
; }2 Q& t, I: h! j4 e2 G
6 X3 k5 Q: _" \%新数据仿真
3 m, h6 ~& T- |9 R& a* ~, V* bpnew=[73.39,75.55
    3.9635,4.0975
% f, g& x$ r& D1 N5 n$ S. E! e+ ~* g6 P    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
% g% @" R9 d4 {& U! C" |/ L' zanewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
6 d, E2 J5 X- Y, |9 h2 v（修改的地方用颜色标记了）
/ l: ]/ A2 ~) N& |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
   \" S% V4 D5 J* r3 ~0 D) h
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,...; C' z/ ?+ U3 r! V0 t
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   , Y) O- ]- q. ?( T- ]7 l* [# ]
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,...: V# ?% r/ I3 G7 W  N0 Q% [9 Z
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   2 ^9 m. ]; ~\" ^; a7 h
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,...
   # O3 I$ Q8 E- z0 x5 l/ V% b
7.     0.56,0.59,0.59,0.67,0.69,0.79];\" t3 u$ x1 |' I: V3 B
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   / @\" f1 x9 o  a\" e) @0 H
9.      22598,25107,33442,36836,40548,42927,43462];
   , k# @4 t# k. ], y
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    , ]5 y2 p( l6 ^% n1 q
11.      13320,16762,18673,20724,20803,21804];
    + j1 b0 F- _( ^* M3 v\" t( {6 a
12. p=[sqrs;sqjdcs;sqglmj];  h, t# A3 q3 R* k2 p5 c
    
13. t=[glkyl;glhyl];
    : B1 y+ N/ X2 @: D6 a0 \
14. %数据归一化
    , y* F9 O* }; T. H7 d
15. [pn,ps1]=mapminmax(p);7 t( B6 H) n\" b3 O0 k* q- l
    
16. [tn,ps2]=mapminmax(t);5 S4 R: X& d; J0 ^: m7 U; }
    
17. %dx=[-1,1;-1,1;-1,1];. X4 t+ b0 X6 s. [2 c
    
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]0 z+ P7 X( p$ A
    
19. 5 O9 n6 S- q9 A' P; \: o2 {9 S' p
    
20. %BP网络训练; V; ^/ w& m0 y
    
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]8 n: O2 g6 @7 M3 T. s
    
22. net.trainParam.show=1000;/ j! U- B. f6 c2 G
    
23. net.trainParam.Lr=0.05;; h6 i2 p) A3 _& w7 x
    
24. net.trainParam.epochs=50000;
    4 X& k/ x, Q' Q\" }) z4 I
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]
    . b# o8 y  L# ~1 I) c2 s% n$ V( ~
26. net=train(net,pn,tn);2 H\" K: k8 F; l! R9 k: w\" W
    
27. 
    5 `; o0 ^# v- n) G\" T8 t
28. %利用原数据对BP网络仿真
    + u+ f! O2 W+ R5 r& Z
29. an=sim(net,pn);
      F, ~* U8 _4 O6 h- S% @6 r
30. a=mapminmax('reverse',an,ps2);
    4 C+ u: G1 R- O$ j\" E7 J/ M# Z
31. 
    8 _( W9 ^. P% G( t
32. %仿真结果与原数据对比测试
    ! Q9 ^* F' G- Z( U
33. x=1990:2009;, e7 L/ e% E3 c* v( T5 \! ^
    
34. newk=a(1,:);
    + d& o+ h# ^8 |: k+ ]
35. newh=a(2,:);  [, [; Y/ \- q! z* g
    
36. figure(2);, `. i9 s6 k( q7 f% q
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    \" M! v4 [# b5 ]' S4 f; O7 d
38. legend('网络输出客运量','实际客运量');# u( x) q2 l! n! W  N1 j6 A1 b5 I7 R
    
39. xlabel('年份');ylabel('货运量、万人');
    : n3 z8 j  C5 n9 y1 R/ X7 e
40. title('运用工具箱客运量学习和测试对比图');
    : i$ |0 m9 _( d
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');. L1 }% U\" D' c9 y9 M
    
42. legend('网络输出货运量','实际货运量');: {\" s+ x; r  _* |  v
    
43. xlabel('年份');ylabel('货运量、万吨');
    ) [, s4 C  `5 f/ M2 o  L
44. title('运用工具箱货运量学习和测试对比图');
    \" |3 F2 A- b' ?' I/ b. x
45. + Z6 w& b+ D5 Y; A. `. ]\" n6 j5 p; o
    
46. %新数据仿真
    9 t  X- j+ Z( C' S5 v
47. pnew=[73.39,75.55
    8 S, N5 |( T& R' w0 i4 F4 ^9 r6 \
48.      3.9635,4.0975, p7 V4 O0 h: O# @  d/ j) E; G
    
49.      0.9880,1.0268];( F* n+ Z8 L6 q9 q3 l7 q# Q7 Q
    
50. pnewn=mapminmax('apply',pnew,ps1);, A* N, l* @3 ^# T7 O
    
51. anewn=sim(net,pnewn);
    5 f. a& n( i) V/ t3 @. ?3 b5 G
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. clc2 u8 q8 k& W2 }* 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,...
   $ g\" N; m9 H8 T: H* N2 f5 B; ]
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];: G6 J0 Z, d! i* N$ S
   
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,...
   4 ^! B) C6 {( m2 d! }) y\" S0 s
5.     2.5,2.6,2.7,2.85,2.95,3.10];# Q: L' W; `3 p\" n: n
   
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 s7 K; ^! o+ [
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   1 _+ q2 d7 P- G2 A: D8 Q\" g/ _' v
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   3 M' ?\" H7 }8 r. C5 \
9.      22598,25107,33442,36836,40548,42927,43462];8 J# m( ^6 ]: w7 I
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    % G& a6 Q7 ^' F* S! `+ l! z
11.      13320,16762,18673,20724,20803,21804];8 g1 {6 [0 ^8 T: y+ s- T1 L
    
12. p=[sqrs;sqjdcs;sqglmj];+ \( @+ r# o' j3 y, {6 X# B) R
    
13. t=[glkyl;glhyl];
    & M. g. v: t6 \% a8 o
14. %数据归一化
    ' V) P! ~\" M$ U) g
15. [pn,ps1]=mapminmax(p);
    2 ~: K% }: u/ m- M( i
16. [tn,ps2]=mapminmax(t);
    4 ^: z. J9 _; Z8 \; F
17. %dx=[-1,1;-1,1;-1,1];\" o1 b' o9 L9 `7 s$ o& _
    
18. p0=minmax(pn);t0=minmax(tn);+ X; z' c8 P! `
    
19. 
    3 |% a3 a3 [9 x& d# o2 c* o% C( z5 n; v\" b' G
20. %BP网络训练; D0 u1 J! ]\" D. }% U& h
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
    4 R8 X; ~2 R7 L, b1 Z' S: O) T! \9 w
22. net.trainParam.show=1000;
    # U  v% U4 A+ w6 A: d) q
23. net.trainParam.Lr=0.05;8 R; B' @, V; I- H' A# W  [/ I
    
24. net.trainParam.epochs=50000;) f3 o7 o; B5 _& [. f4 P& f, k% L) `
    
25. net.trainParam.goal=0.65*10^(-5);
    - {/ K$ J4 F+ {  W/ o$ \/ b! f8 R
26. net=train(net,pn,tn);
    % ~4 Y- z; p. b& z
27. / H# }. V$ U- B. q
    
28. %利用原数据对BP网络仿真7 E! S+ s' Z$ V7 _. P, r\" [+ x
    
29. an=sim(net,pn);
    6 U9 `/ Z0 r1 T# j: o% q; {
30. a=mapminmax('reverse',an,ps2);7 E. b# X( P; C\" S$ V
    
31. - [: N' t) T4 z- p9 S! [
    
32. %仿真结果与原数据对比测试
    4 x' l* s* a  h\" R+ F
33. x=1990:2009;
    . T! ?7 n; a! u\" W, Z, z
34. newk=a(1,:);3 _: M$ T1 a8 H& ]9 X7 ~) b
    
35. newh=a(2,:);3 X3 ]3 B( ~* G  ~6 @! r1 r
    
36. figure(2);
    \" g3 G# `1 X6 S% D
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');3 d+ ?$ h9 J1 f4 _. F1 Q0 t
    
38. legend('网络输出客运量','实际客运量');; h: p$ |$ e- G% C# W# C- F
    
39. xlabel('年份');ylabel('货运量、万人');2 L9 E  Z$ m8 T: {
    
40. title('运用工具箱客运量学习和测试对比图');3 Q+ v9 a5 o6 R# E
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');  O& }* [& b+ k& L4 p
    
42. legend('网络输出货运量','实际货运量');
    # E3 g3 v5 ~3 z7 B9 p
43. xlabel('年份');ylabel('货运量、万吨');( `5 h+ x. n4 g1 P# c( p7 p
    
44. title('运用工具箱货运量学习和测试对比图');
    ' \# @: x7 X' s7 Y6 E6 u
45. 
    / w) }/ O2 A7 y+ e5 p7 C. s/ o
46. %新数据仿真
    0 c: f. P  J: S  W) _) }- U. e
47. pnew=[73.39,75.55
    . N$ z2 f6 E0 R( A0 x
48.      3.9635,4.0975
    9 i4 {& q3 a5 z9 M8 c% K  D
49.      0.9880,1.0268];
    1 ~4 E( p9 L1 w! x
50. pnewn=mapminmax('apply',pnew,ps1);
    * [7 J! x4 u- c: X) U# o: ?
51. anewn=sim(net,pnewn);, U6 i- w6 {1 x3 a( l1 I/ Z4 m' A. S
    
52. anew=mapminmax('reverse',anewn,ps2)

复制代码

T-Eric

且生 发表于 2014-7-21 18:22
% k$ A' ]4 I9 W# x2 Q+ t8 [忽略我上面那个，改动的地方在18，21，25行……求批

& i- e- T  g, i6 g) l2 ^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8 G$ y7 J2 V) {5 g$ g# q) e2 w
   
2. net=newff(p0,t0,6,{'tansig'},'traingd'); 1 a6 M. R2 L  Z/ g5 Y& ]& L
   
3. net.trainParam.show=1000;
   0 }- ~8 @. }6 T\" ^7 x
4. net.trainParam.Lr=0.05;\" I/ k7 B! C# ^( V& y
   
5. net.trainParam.epochs=2000;
   ( q7 V8 d2 w! ]* y; a( D7 C
6. net.trainParam.goal=0.65*10^(-4);
   4 C2 ], V7 Z$ n% k' @
7. net=train(net,pn,tn);

复制代码

我只改了这里面的，
  _: F3 e7 @, Y- {6 z5 S一、改成单隐含层的，6个节点
0 x1 J/ i: w( u  O二、训练函数改成梯度下降BP算法 traingd
8 H9 n0 ^2 D. U5 H* x  V1 M9 N三、迭代次数改成2000
0 q  }  Q+ X; l+ ?! X上面的参数是自己试的，我也不知道为什么。
( _& P, C# X0 R( 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