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
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,...
! x6 s  t* |, B    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,...
, P1 X( X- l( x    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];
# S2 P6 l2 N0 M1 C3 D# l+ _$ Vglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
: B8 N' e' _9 b% I2 O! b    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];
2 U0 O( h8 j" m* J2 O( Hp=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

%数据归一化
[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');
net.trainParam.show=1000;
0 ~& V' Q5 V) u9 ^! x. Anet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
& ^3 k- L. q7 k5 o+ r& r0 Q/ k
/ u" a( T$ @3 I8 O4 b' O%利用原数据对BP网络仿真
an=sim(net,pn);
1 X6 x) G' k+ I6 i% ~2 s! i1 wa=mapminmax('reverse',an,ps2);
# A8 z! }+ n) H) }9 F: l
%仿真结果与原数据对比测试
x=1990:2009;
$ i; n0 T& |3 qnewk=a(1,:);
newh=a(2,:);
figure(2);
/ }/ M( S3 p6 a' bsubplot(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');
0 _0 s# J( Y! @1 h& I; X# k& dtitle('运用工具箱客运量学习和测试对比图');
! K( k7 n, ?3 f6 i2 s$ ]  n0 h. Ksubplot(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');
9 r0 n$ b( `7 ?; K- S  Qtitle('运用工具箱货运量学习和测试对比图');

%新数据仿真
/ g- A3 C% v% k6 o9 B6 Upnew=[73.39,75.55
4 n: F0 ?+ F/ L% }8 r2 e& X    3.9635,4.0975
    0.9880,1.0268];
8 [& C$ k' v2 r. @! K6 i2 ^pnewn=mapminmax('apply',pnew,ps1);
' c, V% Q3 `2 N/ aanewn=sim(net,pnewn);
5 b5 q5 B5 x! zanew=mapminmax('reverse',anewn,ps2)

2 k3 O5 B7 X0 |% W! E

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：
$ W& M) W. i2 Dnet=newff(dx,[3,7,2]{'tansig','tansig','purelin'},'traingdx');
3 r' l% s7 z& e3 R- i( ?1 Z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：
1 O) f0 O. t4 i/ O) d5 T2 l[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
1 d1 l" [8 ~1 _/ T! C6 g- ?8 ea=postmnmx(an,mint,maxt);
5 H+ t1 K# m! X8 _& G& @/ F% Zpnewn=tramnmx(pnew,minp,maxp);
0 t2 M8 z. N9 C8 I- Wanew=postmnmx(anewn,mint,maxt)
/ D& t# `# a3 G4 `我修改为：
[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
  n7 {5 }: E/ G( [2 m% la=mapminmax('reverse',an,ps2);
' c: `9 J1 z! a$ Bpnewn=mapminmax('apply',pnew,ps1);
( [& m# Y3 O0 ]* P6 D7 K  Q; P/ vanew=mapminmax('reverse',anewn,ps2)

1 V! k; i4 M9 d/ n; e4 I: |# }, }. J原程序为：
7 `, r. X+ E: s/ x; \2 ~%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,...
  i3 G! A% c/ k4 L    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
4 J/ i! a& _+ C; l5 f" Q2 s0 r- I5 {7 `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,...
4 q7 V5 }8 i* b    0.56,0.59,0.59,0.67,0.69,0.79];
" |. \4 f* S! x5 n  f. Aglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
6 u  E) K$ a: n$ {    22598,25107,33442,36836,40548,42927,43462];
6 ]' G" S* o  Z: F1 iglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    13320,16762,18673,20724,20803,21804];
+ Y/ X$ D! E$ z6 F0 ]p=[sqrs;sqjdcs;sqglmj];
3 C& o; ]9 a7 o6 M! }t=[glkyl;glhyl];

%数据归一化
[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
dx=[-1,1;-1,1;-1,1];

" V4 m) H$ O0 m) ?  O0 b%BP网络训练
net=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
0 G6 W# |4 z. M# o+ inet.trainParam.show=1000;
+ `& \$ B4 Z- Z: ~+ o  qnet.trainParam.Lr=0.05;
0 m2 u9 T- e$ J" @3 G! X# K; Qnet.trainParam.epochs=50000;
+ \7 L: G9 Y: u. [4 e6 P' pnet.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
4 [) ]( ?, c' B* B( u$ t
7 J. H) w" r1 h' k( B% x( \, k. Q%利用原数据对BP网络仿真
* ~* r& c  d' ^& Q8 p1 f  ]an=sim(net,pn);
( A+ B  i# C, Ja=postmnmx(an,mint,maxt);
! @' M+ |  L$ b; Q8 O
%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,;
: o* W5 f# `0 `, W/ ]4 M2 cnewh=a(2,;
figure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
  l& U! z4 q4 e$ ~legend('网络输出客运量','实际客运量');
4 y8 c5 d! {9 I" ]9 a* o9 V+ e2 \3 t) Dxlabel('年份');ylabel('货运量、万人');
; b/ `9 p' Q; L: otitle('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 U; F9 t  L: y+ ~! e- @" Jlegend('网络输出货运量','实际货运量');
xlabel('年份');ylabel('货运量、万吨');
: s# \. i  u4 v7 u, I  p' ptitle('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
- `' J) }% x2 U+ m0 s$ ?1 p    0.9880,1.0268];
pnewn=tramnmx(pnew,minp,maxp);
8 n# b4 m8 p2 s7 Tanewn=sim(net,pnewn);
# \: ~1 S6 A6 _# `: H1 ~anew=postmnmx(anewn,mint,maxt)

& ]7 J: i' ]( M/ G4 |7 z修改后程序为：
& q# x( K% c) D% S7 m2 @%原始数据输入
  A1 f  @8 I9 P8 O) nclc
  ~4 w9 p4 I1 x7 P( W; Xsqrs=[20.55,22.44,25.73,27.13,29.45,30.1,30.96,34.06,36.42,38.09,39.13,39.99,...
0 |2 ]1 v% ^+ q1 M* p6 B    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,...
- Q7 R6 y# v  R( u3 B7 D    2.5,2.6,2.7,2.85,2.95,3.10];
- N4 x& _4 X: P& P6 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,...
    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,...
6 P  H% l5 w3 G( J# a. n    13320,16762,18673,20724,20803,21804];
% I( u) O+ L& O4 ?6 T9 S8 K+ `p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];
. {) }( o1 I7 B, z
  o  h: G! `! l3 n%数据归一化
[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);

3 H; \0 ?/ V% t. ~5 x# {%BP网络训练
: T3 F# _9 j# T$ m$ t& A  V: p( knet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
, m( U/ n  s/ ]net.trainParam.show=1000;
- w5 N6 |5 T2 W5 dnet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
; |& ~- y- `3 r! c
%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
9 s$ I1 b$ c* i. _% c7 s1 S8 d, zx=1990:2009;
# ^% A, B3 A0 C+ v9 l, Anewk=a(1,;
; n: K5 ^- z, b* M. M' ~newh=a(2,;
- O% E0 L8 o2 g8 j8 F0 [/ J: v# Vfigure(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');
# s7 t. |( i2 P* t; ?. p! S# N5 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');
xlabel('年份');ylabel('货运量、万吨');
+ A  r: |( A9 r8 F! ktitle('运用工具箱货运量学习和测试对比图');

" k4 k5 J# W( T1 C: a9 I%新数据仿真
pnew=[73.39,75.55
    3.9635,4.0975
. M: d0 z) w& h" Y    0.9880,1.0268];
5 E: g% N& q$ U. f  P' vpnewn=mapminmax('apply',pnew,ps1);
* M# o9 j) e% p- [( I# C, qanewn=sim(net,pnewn);
4 y! n( _8 r  L' u# nanew=mapminmax('reverse',anewn,ps2)
4 o* W. h# ~$ \8 o（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: D( @/ }& y5 {4 v  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,...5 I0 j8 M' |. g- J. A) T* J2 n6 e
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];* e$ {: x9 |\" 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,...' B- A4 m, X3 {( O' O) ?% Q. D
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   5 r6 Y3 W, e% `9 K1 d  W
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,...
   : w\" [2 ^0 I8 p+ }\" M
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   ( O& P  `9 H) E) g
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...2 g$ [- L' m, R, T! O( [2 J- B0 f
   
9.      22598,25107,33442,36836,40548,42927,43462];
   ' }5 J3 E5 P3 x* u1 C7 Q7 m
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    ! E* ?0 W. |+ s- c/ k9 z
11.      13320,16762,18673,20724,20803,21804];
    ! v\" i0 V4 [  f/ F7 E% J
12. p=[sqrs;sqjdcs;sqglmj];( Q7 b* m4 L  F
    
13. t=[glkyl;glhyl];
    & t1 w% S% S  m; }% F. {
14. %数据归一化7 z- K8 @9 O4 ]+ h
    
15. [pn,ps1]=mapminmax(p);
    7 i\" }: u0 Y2 Y5 w
16. [tn,ps2]=mapminmax(t);
    , u7 C% S/ q\" `0 P0 ?
17. %dx=[-1,1;-1,1;-1,1];7 l' A\" y1 \3 ^) O& m
    
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]9 A- ^5 R- v5 G7 w3 U% D
    
19. 5 F; m. p. ?% l1 F4 N7 a! }
    
20. %BP网络训练
    : K- P& B8 S6 B5 k; W\" ~9 A: X1 i
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]
    \" G, r6 Y% v4 W+ P
22. net.trainParam.show=1000;
    . T8 n( x$ c+ a; E& {# `0 D$ Y
23. net.trainParam.Lr=0.05;9 I: N# b: D- M( l6 C
    
24. net.trainParam.epochs=50000;
    0 j6 |+ t' G; U$ o& {
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]+ Q6 k7 n2 C( L# l, c+ a
    
26. net=train(net,pn,tn);
    : h# q7 d5 J  s& u
27. 
    ! q& p6 y* N: ^% [1 H
28. %利用原数据对BP网络仿真
    ' a. o: O: l$ n2 W% n& ~
29. an=sim(net,pn);+ I% k& w3 e0 _4 n7 U3 Y# j
    
30. a=mapminmax('reverse',an,ps2);3 B( c\" k1 \' Y8 H' D8 y4 ]. f1 r
    
31. 
    : C3 H; u\" ^# l% ^7 K
32. %仿真结果与原数据对比测试 \" `4 _, N6 M1 q+ k: d
    
33. x=1990:2009;1 `, T/ E' G* I8 b7 X# g# M
    
34. newk=a(1,:);- x# B; v$ d* ?9 e, K
    
35. newh=a(2,:);: u3 L1 g- a1 z  d$ t0 b1 Y% Q
    
36. figure(2);/ Z; H+ Y5 b9 W5 ]% W- a1 G
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');+ ^' b  U/ r: y) V5 B+ E* |% d
    
38. legend('网络输出客运量','实际客运量');+ w! I\" l* {& T1 S# `% r2 ]
    
39. xlabel('年份');ylabel('货运量、万人');, v: w) R. P* c3 [  q+ H8 V2 B
    
40. title('运用工具箱客运量学习和测试对比图');/ s7 Y9 K! c0 ]' p- E- f
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');2 b1 I3 w1 U$ @1 Y
    
42. legend('网络输出货运量','实际货运量');4 n; K: z. z2 K% ]2 l
    
43. xlabel('年份');ylabel('货运量、万吨');
    2 Z+ U  y: s  `$ L- g4 G! {2 s
44. title('运用工具箱货运量学习和测试对比图');4 ]. B- w7 I* r6 ?# e4 \+ Q# m/ [
    
45. 1 G3 r6 G3 ~6 D* @: {3 s1 r+ d
    
46. %新数据仿真
    7 g6 W# L5 X5 A- o7 h
47. pnew=[73.39,75.557 o7 J1 Q- x, D) j+ p
    
48.      3.9635,4.0975
    ! l0 P0 K  O/ ~
49.      0.9880,1.0268];
    . W% ~; I8 p1 P! V8 C$ w/ d9 w9 s; [
50. pnewn=mapminmax('apply',pnew,ps1);
    5 K: @% s, |1 A* l1 h1 e& c* x
51. anewn=sim(net,pnewn);
    8 t4 J& S0 W4 V
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
   . X, L2 h1 {2 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,...
   9 z1 A/ q/ c9 c* K\" \! s* H8 {1 V
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];9 i* l- B- |3 L& d& f1 I0 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,...( ~0 W* m, E& q# L- v
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];, O( V, D1 B# ?4 c
   
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,...
   $ s: c1 K* Z5 y# u+ \8 e! y
7.     0.56,0.59,0.59,0.67,0.69,0.79];( v; a; t/ \\" [$ x+ l\" S\" L, a8 l; H- _
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,.../ J  A  `, K/ \
   
9.      22598,25107,33442,36836,40548,42927,43462];
   & d2 `5 R3 T, V  Q) w8 E
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...; y2 `; C( Y  u
    
11.      13320,16762,18673,20724,20803,21804];
    0 f' c( b+ {0 n3 L5 N+ o: o- i: Q\" U
12. p=[sqrs;sqjdcs;sqglmj];4 C% i2 O8 i: v, y! S: |( B* K
    
13. t=[glkyl;glhyl];
    - M+ d  |2 ^- u8 \
14. %数据归一化
    ; n) @/ N1 ^/ ~
15. [pn,ps1]=mapminmax(p);1 y4 L  q- N7 x8 P
    
16. [tn,ps2]=mapminmax(t);
    . E% J( t7 T8 U; Q' n- ~  w# h
17. %dx=[-1,1;-1,1;-1,1];
    ' y0 G8 T1 }( d4 `. ?\" P4 D- ^
18. p0=minmax(pn);t0=minmax(tn);$ ~# [+ _4 P) o
    
19. 
    7 h+ K* j# @+ I! O- `
20. %BP网络训练
    : f. ]4 \7 ^- ]- x0 v) l+ l
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); \" l8 J- K3 L$ F
    
22. net.trainParam.show=1000;3 \% ?0 z' w7 L$ l  T9 N. [1 X  i
    
23. net.trainParam.Lr=0.05;
    0 |; {4 T: F* n% H( G
24. net.trainParam.epochs=50000;
    4 Q: S- c$ k+ Q3 a2 c\" t\" ?! A. |
25. net.trainParam.goal=0.65*10^(-5);
      Z+ \9 K6 d3 e: Y  J3 X9 Z) v) m
26. net=train(net,pn,tn);
    * }2 C- h3 L$ b, ^9 P. G  A* H2 B* c# }
27. 
    , R5 J. l% _3 d4 l+ K. S
28. %利用原数据对BP网络仿真% N; [6 C) f% y
    
29. an=sim(net,pn);
    ; z. x) G3 J. a) w6 k& g\" \7 r
30. a=mapminmax('reverse',an,ps2);, K* K  }- {& @4 ]2 i
    
31. 
    5 s9 J( V( q. F! I5 D& T+ A
32. %仿真结果与原数据对比测试
    1 r6 L$ @/ i' e- i
33. x=1990:2009;
    ! q' \$ O9 X' X3 w  G
34. newk=a(1,:);, S' j9 _\" _% N- n1 k6 K
    
35. newh=a(2,:);
    % T2 n- _) J2 m# j
36. figure(2);
    4 l6 e$ z; @* u
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');5 n0 A! s+ W9 q$ S6 ?
    
38. legend('网络输出客运量','实际客运量');
    , g' |' G9 a8 J
39. xlabel('年份');ylabel('货运量、万人');
    . _; z. x\" D' g2 `! b3 I$ F
40. title('运用工具箱客运量学习和测试对比图');- B& S\" \9 x) n9 K- K
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');# j6 b\" H9 e6 _) F\" X' J& x; E
    
42. legend('网络输出货运量','实际货运量');
    / a! R8 ~4 c1 l) K# ]
43. xlabel('年份');ylabel('货运量、万吨');
    ) j\" t' a9 D8 Q, ?+ F1 m
44. title('运用工具箱货运量学习和测试对比图');7 S$ o\" _7 w( r& y
    
45. 7 C8 f# v. f: m* G3 u  I3 a( F; ]
    
46. %新数据仿真
    7 N2 t\" R; J7 j- e) i9 o6 c
47. pnew=[73.39,75.55
    % w9 c: X) J( d, Q, J! z6 W+ |
48.      3.9635,4.09756 w- O7 w* r; t5 G\" T
    
49.      0.9880,1.0268];
    7 i/ A0 J' R) S' i' ~; m
50. pnewn=mapminmax('apply',pnew,ps1);
    6 v* [) y9 R4 l9 {2 O
51. anewn=sim(net,pnewn);
    ; s& ?0 @1 \5 r; q
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

7 Z/ i/ J4 p# e. h0 n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
   : r% N8 t0 Y+ l% n* H4 M3 T# u
2. net=newff(p0,t0,6,{'tansig'},'traingd'); 8 X  g; a4 i) x2 O4 h
   
3. net.trainParam.show=1000;# F) i1 ^( O1 I3 H1 A* E- f0 Y6 Y
   
4. net.trainParam.Lr=0.05;% o! X- \. @1 c# D
   
5. net.trainParam.epochs=2000;
   : v. D7 r: ]\" u  `! a; j, l
6. net.trainParam.goal=0.65*10^(-4);% k$ W- x* l, d& g9 v
   
7. net=train(net,pn,tn);

复制代码

我只改了这里面的，
9 r5 e7 b5 e7 F一、改成单隐含层的，6个节点
二、训练函数改成梯度下降BP算法 traingd
+ P1 k$ S: f! H3 x三、迭代次数改成2000
% s: q$ C, R1 D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。
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