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！！

/ c0 m5 U+ }2 g, N%原始数据输入
clc
3 L7 [) i. E" ?6 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,...
    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];
0 g+ _; l# O% zsqglmj=[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,...
; S5 t$ x7 B. \! x7 z7 z7 q1 l    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];
; f' ]9 c& m5 D& S8 W, M4 B& O# O2 Tglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    13320,16762,18673,20724,20803,21804];
! ~* q/ V: @) V  L& y5 V4 Jp=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

%数据归一化
$ J% v# |. p1 v3 u" g0 `[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
/ O  S9 Z+ U* G. m3 fdx=[-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;
, J' x- {6 g7 w" ^8 Z/ b6 onet.trainParam.Lr=0.05;
- v/ \; X; ?1 ^net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
& s7 d( q& l' [, }# A3 Bnet=train(net,pn,tn);
1 [0 D, {; c. ]" p, ~  n
6 \* q6 q3 J2 I) l2 w( x/ R6 R%利用原数据对BP网络仿真
an=sim(net,pn);
a=mapminmax('reverse',an,ps2);

# i3 t  a9 F# |( \. q1 n%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,:);
4 ^6 C# A0 o( g5 ynewh=a(2,:);
figure(2);
7 F% M& P! i9 V! ]! xsubplot(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');
3 ^4 e- B0 S6 K6 h, K# D9 i+ M6 e- rtitle('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');
, `& D/ z: ]- y5 g( ~" Mtitle('运用工具箱货运量学习和测试对比图');

%新数据仿真
4 u% g4 ?- x: X7 [$ K6 R0 cpnew=[73.39,75.55
    3.9635,4.0975
    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)

* l6 i4 T- z* i  A1 C! P; R

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
* m  k6 b5 J% S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：
net=newff(dx,[3,7,2]{'tansig','tansig','purelin'},'traingdx');
我修改为：
8 u5 \% m+ e6 u' t4 vnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
9 w) z# n6 l3 I" {3 t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：
$ r  L4 Q/ v6 z" b  J3 r7 ]+ t[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
a=postmnmx(an,mint,maxt);
8 ~! \8 U. q1 k6 o2 O3 i1 Z7 Y" Rpnewn=tramnmx(pnew,minp,maxp);
anew=postmnmx(anewn,mint,maxt)
& M4 {+ M- s2 j1 u, U/ R我修改为：
[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
a=mapminmax('reverse',an,ps2);
+ p! `% W! }5 g" F) kpnewn=mapminmax('apply',pnew,ps1);
anew=mapminmax('reverse',anewn,ps2)

3 C* k; r" W/ {/ n: Y" y原程序为：
%原始数据输入
clc
# x' \4 S$ y( 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];
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,...
8 i$ U% K; A9 n& J9 R. r+ b    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,...
- f4 K0 {" x0 X2 S. W    22598,25107,33442,36836,40548,42927,43462];
9 x1 g9 P7 J# Sglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
, c3 n; a% h$ e3 wt=[glkyl;glhyl];

%数据归一化
( g7 V  d& r; C9 \: U[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
dx=[-1,1;-1,1;-1,1];

' P6 Q8 i/ `% U1 `) b0 Y1 p) P%BP网络训练
1 W- j, N  C9 N) i. k6 Enet=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
net.trainParam.Lr=0.05;
" A0 a1 v* b* t* J4 C9 `net.trainParam.epochs=50000;
8 b: f) b# B& X& L- \) W1 G2 Hnet.trainParam.goal=0.65*10^(-3);
2 f- J+ V) }6 U4 ~& n9 cnet=train(net,pn,tn);
0 g6 c( F  C: {; ^3 N+ k& ?
7 o  A( E  g( m7 K& h$ K%利用原数据对BP网络仿真
an=sim(net,pn);
a=postmnmx(an,mint,maxt);

( N5 z# a1 Q5 r1 I%仿真结果与原数据对比测试
x=1990:2009;
; F, ^3 l9 m2 g$ N! Qnewk=a(1,;
newh=a(2,;
figure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
- b: M6 o* F/ D, m$ ~legend('网络输出客运量','实际客运量');
% R! M8 K  }* y" Wxlabel('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');
* a* {9 K& Y% {subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
# r$ v+ N# k' y+ [& r9 U0 ^legend('网络输出货运量','实际货运量');
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');
( ~! T; u  i+ x5 Z  p
%新数据仿真
pnew=[73.39,75.55
4 n2 p' r, F, U! n    3.9635,4.0975
    0.9880,1.0268];
& b2 L3 U0 `  e7 K) O3 u2 q& Tpnewn=tramnmx(pnew,minp,maxp);
1 K3 b7 M9 a  K6 s' Ianewn=sim(net,pnewn);
anew=postmnmx(anewn,mint,maxt)

修改后程序为：
%原始数据输入
' Z& @' _5 ~- r/ H, \  Pclc
5 t* p. h) Y  _! d4 S0 H6 Y, ssqrs=[20.55,22.44,25.73,27.13,29.45,30.1,30.96,34.06,36.42,38.09,39.13,39.99,...
+ X2 s: v2 [7 N3 n- W0 d    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
2 A& ~" F7 B+ Z" H) wsqjdcs=[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];
5 q+ G; A( M9 l. Y' usqglmj=[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 S) }! ~% u" M/ y- f3 n0 \+ @! |    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,...
" e0 Q6 ~3 G7 t( T) d    22598,25107,33442,36836,40548,42927,43462];
7 V9 ^2 e* G+ s) v9 W$ yglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
$ \- B4 q- f  z, V3 f% D5 E. G    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

6 \+ u4 j! ]& P$ L8 q%数据归一化
[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
, C7 O* \$ j( o0 A) D% s+ ]9 y4 v5 J
$ u) A( c% \  c! W. c$ V7 f5 ]%BP网络训练
: `7 V0 p  Z' A' T9 S% w3 U8 Anet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
net.trainParam.Lr=0.05;
. R0 Q# y  @& Vnet.trainParam.epochs=50000;
' o9 ?3 L* v* y: @8 a( Wnet.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);

8 }9 M; n7 d9 ?7 m$ s1 U. |# h( P" h%利用原数据对BP网络仿真
an=sim(net,pn);
a=mapminmax('reverse',an,ps2);
3 J# Q; r; r: i  b2 f* g
8 k4 m3 T4 H$ z, Y! ^%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,;
newh=a(2,;
figure(2);
& P3 @- ^$ u  Q  x: Bsubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
. m4 R$ a) }. t6 u9 j( e  ulegend('网络输出客运量','实际客运量');
# B) k. h2 U- S3 Xxlabel('年份');ylabel('货运量、万人');
$ J; h: |1 j* O/ }/ d$ R' 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');
title('运用工具箱货运量学习和测试对比图');

* s& [, q1 L7 d1 Q0 ~%新数据仿真
pnew=[73.39,75.55
, d$ d" J. B5 g; j    3.9635,4.0975
) M5 B- A4 }+ F3 I! D8 b5 E    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
- c" \  C- z2 z5 Y. Danewn=sim(net,pnewn);
. u% ^! f. B$ z: v: Vanew=mapminmax('reverse',anewn,ps2)
（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
   % ~7 N\" D1 U9 j
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,...
   * A' u3 ?2 Q5 A7 q& O8 p* B
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   1 m2 W3 |4 e2 Y# S* f' b! V
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/ ^! W  F/ f% t' F
5.     2.5,2.6,2.7,2.85,2.95,3.10];6 O% T9 n' \8 Q6 k, g, Y( 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,...$ p6 h\" a8 f/ b8 v7 U
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];. H* ?  ]( ?1 r$ e
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...  L. I& [  U/ a3 j! Q# r
   
9.      22598,25107,33442,36836,40548,42927,43462];
   4 T1 _) u/ g% a
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...  x; v1 Y\" g& D& z) q' A# R
    
11.      13320,16762,18673,20724,20803,21804];/ ]2 A/ [; k7 H0 p8 }! [: q; A
    
12. p=[sqrs;sqjdcs;sqglmj];
    & D1 J6 [3 A0 ~
13. t=[glkyl;glhyl];
    7 B$ D\" y) X( }2 |( c  v$ {
14. %数据归一化8 G  Y1 ]4 t* \3 z9 V6 Q' z/ v; E
    
15. [pn,ps1]=mapminmax(p);
    6 Y2 D( m- Z; f7 z! E: V2 G& q
16. [tn,ps2]=mapminmax(t);
    $ Z+ h4 [! D6 K9 m$ q! l3 g\" W: V
17. %dx=[-1,1;-1,1;-1,1];7 p- F* {3 |' w6 n  E4 c
    
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]6 `# c* n+ k/ u# Y  `! A& m4 n
    
19. 
    % m( v* h- Q6 m3 m# p/ H* P# j
20. %BP网络训练* z' z8 E; a4 @3 h7 r7 B9 h
    
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]4 D& e: n( r( n% J8 Q! D
    
22. net.trainParam.show=1000;
    4 g8 h5 Y9 F9 b5 T) f
23. net.trainParam.Lr=0.05;( k% W' L5 v3 z) ~3 i
    
24. net.trainParam.epochs=50000;
    ' Y  J: G$ ~- \
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]2 x% H8 w0 z0 J$ w- W7 [
    
26. net=train(net,pn,tn);
    7 |4 Y1 x\" K7 S& P/ a3 f0 K5 ]
27. 
    \" d) y; `& ^\" Y! W8 L
28. %利用原数据对BP网络仿真
    . a6 F3 j8 }7 ~1 q6 U9 |% q6 u. X) W
29. an=sim(net,pn);
    . p8 X9 m9 `6 n7 f1 }% p6 V
30. a=mapminmax('reverse',an,ps2);# @; x5 R% l$ N( V1 q6 [  ]
    
31. 9 u1 F' Y/ z7 g6 i7 ]  a. x/ q# r
    
32. %仿真结果与原数据对比测试 \" r+ L7 `4 N8 d- L) J4 z
    
33. x=1990:2009;5 G0 p9 e  V8 \* s9 x
    
34. newk=a(1,:);\" s- D- R\" Z: \% V3 X# s
    
35. newh=a(2,:);
    . p  k- _/ f) h/ \% r4 M( W
36. figure(2);7 Z+ U+ g- a: a& P* |3 `: k
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');- O% e+ S\" j\" c7 d\" x% F! ]
    
38. legend('网络输出客运量','实际客运量');
    + W% d6 ?+ h  c) b
39. xlabel('年份');ylabel('货运量、万人');
    4 f. Q& u; l% w/ K# e# e
40. title('运用工具箱客运量学习和测试对比图');
    2 O6 ?' B' F2 z. {
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');* j\" f- b' V( Q/ ~% x* a( Y6 t# I
    
42. legend('网络输出货运量','实际货运量');
    7 w# y7 _0 e5 X- n( {
43. xlabel('年份');ylabel('货运量、万吨');- @4 o5 e) h1 ?- f' \
    
44. title('运用工具箱货运量学习和测试对比图');% Q; ]6 T. V4 K7 E! c
    
45. ! l+ I4 [$ C* Y\" y, u) E
    
46. %新数据仿真
    + J, E/ T$ f, a
47. pnew=[73.39,75.55# d* J* O\" @. S+ j; r
    
48.      3.9635,4.0975
    7 z, y2 Z  W2 n. ?( z
49.      0.9880,1.0268];
    3 m; Z$ d+ b( K& v3 ?' {  {
50. pnewn=mapminmax('apply',pnew,ps1);
    / {2 |7 D5 f2 A\" b' F& z
51. anewn=sim(net,pnewn);) w\" y' L3 O! _' W  F' p. c
    
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$ ^2 w- _/ w5 {6 a
   
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 f% g% N4 w/ U3 ?1 }) {3 n
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];/ m0 T. V\" n8 H/ a2 ^1 n% V' _8 H
   
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,...8 c- t% v; y+ b- c. n% D# C; o7 L
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];( J7 U/ c$ p. 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,...
   6 Z$ f5 x9 A3 l& X, Q2 o
7.     0.56,0.59,0.59,0.67,0.69,0.79];7 a/ x3 X7 s$ z8 _& E' G% A0 c) V3 V
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   1 }2 Y! a7 H9 `' N2 {: E/ K. R. E
9.      22598,25107,33442,36836,40548,42927,43462];
   , X7 Y- Z; c$ x# V9 e8 n
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...& {! c7 K, {$ u8 x\" n  }
    
11.      13320,16762,18673,20724,20803,21804];
    * B0 g3 {) T4 K) R: Y2 |
12. p=[sqrs;sqjdcs;sqglmj];
    & G+ l6 ]' v8 h! ?1 e7 S
13. t=[glkyl;glhyl];& ~8 b, K3 i\" u. d, o2 ~  Y
    
14. %数据归一化; `. f8 [2 m' p
    
15. [pn,ps1]=mapminmax(p);
    ; Z, Q. t4 }/ w+ M# Q
16. [tn,ps2]=mapminmax(t);. ^3 t: ~' P\" j6 S\" V8 \! A
    
17. %dx=[-1,1;-1,1;-1,1];
    0 K' l' w; f5 U( [' }2 \1 n
18. p0=minmax(pn);t0=minmax(tn);
    0 T! S8 R1 q\" |- V\" M3 f* D! a! a6 g
19. 3 P) g. p8 i% L
    
20. %BP网络训练
    , F! C5 x/ [3 L5 `' e' h
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
    + B6 i) K5 y5 P+ I
22. net.trainParam.show=1000;9 a0 O3 W4 L* m2 v; K3 W0 L
    
23. net.trainParam.Lr=0.05;6 r7 W6 A/ `) k6 N7 [0 ]- u( z
    
24. net.trainParam.epochs=50000;
    1 K% z$ V- U0 j' B: E
25. net.trainParam.goal=0.65*10^(-5);7 w1 V\" y% s$ J) j8 d2 L: O
    
26. net=train(net,pn,tn);; |: R8 V2 A. O# r$ V
    
27. 6 J) a! Q# X8 S+ @& j
    
28. %利用原数据对BP网络仿真
    + y2 b- ?9 ]# [, m6 m
29. an=sim(net,pn);# B\" v+ D. y4 O% N) J4 i+ g\" S
    
30. a=mapminmax('reverse',an,ps2);
    5 T& Z7 a5 f; d4 y! u, T# V( X
31. 
    6 a7 C0 ~& B* I( b) _7 M
32. %仿真结果与原数据对比测试
    ( @: O& W8 i, O$ T\" m
33. x=1990:2009;1 z* u: @$ }$ s+ \& `8 g9 v. K
    
34. newk=a(1,:);
    5 ~- S4 }. p# R! J
35. newh=a(2,:);
    # C2 t. [6 \1 \: P, W2 t2 x\" \, s7 `
36. figure(2);
    # e7 K- q% ?- O+ L- g* e
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');2 Z! [9 E' p6 H& ^% T
    
38. legend('网络输出客运量','实际客运量');3 r' K; }4 P* K- n6 l4 O1 Z4 m
    
39. xlabel('年份');ylabel('货运量、万人');
    6 h/ m% l& ]! r; K# }
40. title('运用工具箱客运量学习和测试对比图');
    3 b* f/ h* H* F3 B* Q  k
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');5 [6 a& V  J9 p- g- U
    
42. legend('网络输出货运量','实际货运量');
    6 s4 _* b4 W' Q2 ~% b
43. xlabel('年份');ylabel('货运量、万吨');
    3 M1 i0 Q# z$ C% N$ _4 z0 H
44. title('运用工具箱货运量学习和测试对比图');
    \" r# P. @9 h8 d! v- j- H3 }
45. ! B; G\" C1 j$ d4 \! m
    
46. %新数据仿真
    8 J0 K) E8 {% v0 J1 Z, q9 x
47. pnew=[73.39,75.55
    9 M8 F/ D, Q) q5 b, G
48.      3.9635,4.09754 T, O\" m# y* C3 b/ a4 Q4 ~
    
49.      0.9880,1.0268];
    4 a\" N2 ?* x- C1 B
50. pnewn=mapminmax('apply',pnew,ps1);
    & U  a8 b* ~; H  ~\" y2 i+ I$ o
51. anewn=sim(net,pnewn);3 @# @' H+ M$ w\" ^( _$ q( t\" [
    
52. anew=mapminmax('reverse',anewn,ps2)

复制代码

T-Eric

且生 发表于 2014-7-21 18:22
8 j# Q. }) p* O7 h! ^8 p* z* V$ C6 P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4 e- ^9 u* O\" ^
   
2. net=newff(p0,t0,6,{'tansig'},'traingd'); & l& w6 P( R9 i- @
   
3. net.trainParam.show=1000;
     y& @* V' h# @) t& B# `
4. net.trainParam.Lr=0.05;
   $ N, `/ A- Y! v
5. net.trainParam.epochs=2000;
   - E. w\" K3 v* ?\" J' s  r
6. net.trainParam.goal=0.65*10^(-4);8 L( J2 G0 `: [, {( C$ k
   
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
7 a3 F; {1 @3 ~0 P3 I6 D' ^二、训练函数改成梯度下降BP算法 traingd
三、迭代次数改成2000
- f+ _4 p# \$ p0 _8 }) L' B上面的参数是自己试的，我也不知道为什么。
  t" I$ ?/ n- T- m3 p& H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。
7 m* t1 g% L7 ]0 ~6 x3 X3 p( t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