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！！
9 w1 S/ |+ [+ ?7 f# D: \
%原始数据输入
! [! |" W" u, Vclc
! V1 E" X* S4 a% b5 [" g0 zsqrs=[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];
9 I3 X( v9 s& S! e% Vsqjdcs=[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,...
    0.56,0.59,0.59,0.67,0.69,0.79];
' |6 _' _! f/ p- P4 |glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
- w! @& v! `3 ^+ L    22598,25107,33442,36836,40548,42927,43462];
! c; N% ]" g& R6 G/ ?' n  Fglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
- I# j$ v: z* B- _1 [; @1 u    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];
0 F- D1 Q7 O/ u, v! q# y! v3 o
%数据归一化
[pn,ps1]=mapminmax(p);
% z, e6 {6 U& M/ g" y  x[tn,ps2]=mapminmax(t);
dx=[-1,1;-1,1;-1,1];

%BP网络训练
; f2 c7 `1 h* n6 Q; fnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
, d% }8 \5 e" O+ `net.trainParam.show=1000;
net.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
7 M: l. h# N% K+ w3 K7 Q
%利用原数据对BP网络仿真
an=sim(net,pn);
a=mapminmax('reverse',an,ps2);

( b% ]% l8 }3 i) M- p9 j$ \+ A( t%仿真结果与原数据对比测试
2 L( q% q& K9 t) x$ G7 R' tx=1990:2009;
newk=a(1,:);
* u4 M- ^6 j5 z  V" c* `$ [newh=a(2,:);
figure(2);
1 @: k/ k5 J% o7 c0 N; t/ f  \4 ?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');
2 I6 s- L6 _& T. Csubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
" h, a8 Y% P8 Q- g( e6 y) Zlegend('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');
2 R! e$ T- V) z" M8 j$ I- J
%新数据仿真
pnew=[73.39,75.55
7 P( G1 M5 R5 N- X2 ^  t6 S. ^    3.9635,4.0975
    0.9880,1.0268];
7 \0 b2 W9 o6 h1 g. Ppnewn=mapminmax('apply',pnew,ps1);
: d0 h& m) A) y- T9 R! q! Uanewn=sim(net,pnewn);
  U4 s" v# n6 hanew=mapminmax('reverse',anewn,ps2)

) ]& q$ e  s$ O

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
4 |* Z: W& ?( B$ d% G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：
9 f* Y: L. y* i6 Tnet=newff(dx,[3,7,2]{'tansig','tansig','purelin'},'traingdx');
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：
- }  {+ V/ g0 q2 |[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
8 A- n+ Q8 y. o- Y$ ta=postmnmx(an,mint,maxt);
pnewn=tramnmx(pnew,minp,maxp);
anew=postmnmx(anewn,mint,maxt)
0 `6 B: X5 F. d4 a我修改为：
3 n% e! n' \: r0 ?[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
a=mapminmax('reverse',an,ps2);
pnewn=mapminmax('apply',pnew,ps1);
4 D  ~5 v# K5 R7 s; j- \anew=mapminmax('reverse',anewn,ps2)
& a! g9 e! o3 w) B/ r
! J6 ?+ y9 J4 Q, r3 X% y原程序为：
%原始数据输入
clc
3 W& ]3 [& t: f' w' ]) Tsqrs=[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,...
- o, ^0 T. ^2 a2 _3 W; W& G    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,...
. `1 p6 j, o/ E! Q( l# Y5 S- K    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];
5 D+ _$ W+ [: k, ?" v' kglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    13320,16762,18673,20724,20803,21804];
9 K4 d" G3 ]$ r7 v8 R3 mp=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];
  u% q! ]+ D- B; r% Y4 h# P: P
%数据归一化
[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
' \8 u' f1 U4 e, f4 rdx=[-1,1;-1,1;-1,1];

%BP网络训练
net=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
" W  T) v( L8 O1 v* V) ~: }0 q4 A8 onet.trainParam.show=1000;
net.trainParam.Lr=0.05;
. d5 e6 X- B. z0 U% S7 v2 ^0 M6 w4 b. [net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
9 d# R" w2 n+ {: B. x6 y' E4 z# n& n. A
%利用原数据对BP网络仿真
, F* t  Z+ L) g; U6 aan=sim(net,pn);
! v- @9 `. O; y, c+ ia=postmnmx(an,mint,maxt);
* L9 ^4 R& ]7 U$ U) e3 V
%仿真结果与原数据对比测试
5 b0 n- \. h- i2 ~8 Q0 W0 vx=1990:2009;
& }/ j  h  j4 T* Q6 o' ~newk=a(1,;
7 T8 ^; Z2 f9 ?: ?* Anewh=a(2,;
7 ?* }5 W( C, ]( ~figure(2);
9 W4 Q$ [3 X' `+ D! _4 B7 h; wsubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
$ N0 A% Q; |1 t8 Hlegend('网络输出客运量','实际客运量');
xlabel('年份');ylabel('货运量、万人');
4 ]2 O: X2 w, Q* Y# \2 v; o; dtitle('运用工具箱客运量学习和测试对比图');
4 _4 s: m5 X, P* g+ Lsubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
legend('网络输出货运量','实际货运量');
$ y, v1 ?3 u8 |- T7 y  q3 exlabel('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
pnew=[73.39,75.55
    3.9635,4.0975
    0.9880,1.0268];
' n# r# ]3 Q) c! D" u8 S, Fpnewn=tramnmx(pnew,minp,maxp);
anewn=sim(net,pnewn);
1 `8 X7 l* A1 Ganew=postmnmx(anewn,mint,maxt)
  [( I' [1 m8 L  H4 l
修改后程序为：
3 J  e2 X9 a* j3 z$ {+ D, U%原始数据输入
" A+ K% ]( n/ l; d4 l+ ]clc
7 ]- d; s0 H: m% n; Isqrs=[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! O' T0 x3 G* Q! ~0 _    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
5 F$ W0 n4 t. i% p* \% @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,...
5 a* E3 w5 a: e# [: [- j% D3 C    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,...
, A. i# ]7 p8 T! [- {* L6 G6 Z1 i    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];
( u& l3 {/ |) f# `glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
- T; m4 Z. P) v8 l) m    13320,16762,18673,20724,20803,21804];
& a! N2 ?7 }# Rp=[sqrs;sqjdcs;sqglmj];
- s7 |; r0 }; N2 vt=[glkyl;glhyl];
/ t+ ]8 S  y, O. ^% B* P3 Z& ~9 m
%数据归一化
[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);

%BP网络训练
& u8 r; Z6 b7 L# h% f# H  fnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
( u6 I3 C; f+ `: v0 F" Pnet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
+ X* r  D  o( A. gnet.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
( E" Q% B$ m& S2 l4 B
%利用原数据对BP网络仿真
an=sim(net,pn);
a=mapminmax('reverse',an,ps2);
& |& t5 c& J5 d9 V
%仿真结果与原数据对比测试
* G1 I, O; D$ `x=1990:2009;
5 v# Z6 u: f2 }' F9 b: Snewk=a(1,;
  P( K6 B$ w( @1 c7 pnewh=a(2,;
2 _2 S( [4 O; ffigure(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');
1 H3 B0 a) N8 t' I. u; Ktitle('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');
4 B* g' X9 ^" g7 Y) Z8 }) Z6 htitle('运用工具箱货运量学习和测试对比图');

# q, _) A* V, ^! |3 X' M* _%新数据仿真
$ l" D9 w9 Z+ H, apnew=[73.39,75.55
    3.9635,4.0975
2 s$ ?. J4 J9 B& w8 u    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
; j' u6 e6 ?5 eanewn=sim(net,pnewn);
anew=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
   6 x& m: c. i5 C, B8 b
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,...+ x7 i# R# Q* A) H
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];6 I% X5 p- x8 \( {, 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,...$ {: `) q. h  H: d( J\" |; N% }
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   - Z' x$ |, T' P( R% z
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 v8 v9 r! C3 i. P8 B, q9 ^
7.     0.56,0.59,0.59,0.67,0.69,0.79];# S' R/ ^6 Y  B) d2 h
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...' s/ r9 r. _4 `0 z7 ?! Z/ S) G
   
9.      22598,25107,33442,36836,40548,42927,43462];- A, _7 G9 S8 {' `/ A, \. z0 r
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...2 q6 H  ^4 k1 t, y. j
    
11.      13320,16762,18673,20724,20803,21804];
    8 z) `3 S4 I, J/ _
12. p=[sqrs;sqjdcs;sqglmj];6 ]6 f+ Y/ r) y
    
13. t=[glkyl;glhyl];7 ?0 n6 O, u, G: k3 H& Q  i3 U
    
14. %数据归一化: {2 ^0 E4 s$ u2 P8 M
    
15. [pn,ps1]=mapminmax(p);
    4 A5 `- q  a( ]* z6 w
16. [tn,ps2]=mapminmax(t);
    ' B! Q' z* \( D9 v& ]5 d2 q\" p: z$ b
17. %dx=[-1,1;-1,1;-1,1];- t8 j) g; O! p2 S  ^
    
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]
    ' h* V! |; L5 z4 f5 ^2 J
19. 
    / w6 K1 s: D3 O: v9 F
20. %BP网络训练
    2 ~/ y4 f- W; o
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]  B4 a$ @6 x7 ^2 g2 H. [$ z0 ]
    
22. net.trainParam.show=1000;
    : F1 V$ N- w$ C& k0 C9 K
23. net.trainParam.Lr=0.05;
    \" f, P* S; `7 V0 p0 @
24. net.trainParam.epochs=50000;# N! |8 @5 }# b4 n# u
    
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]
    . o% |+ V1 j  t: T
26. net=train(net,pn,tn);, d2 k, O1 N4 j  Z/ A
    
27. 1 {9 r; g) X1 x; j% \) x) P! S
    
28. %利用原数据对BP网络仿真& Y0 j6 ~% X8 a. R% J
    
29. an=sim(net,pn);
    # S  A. y8 U2 L; [' P$ P
30. a=mapminmax('reverse',an,ps2);
    - Y9 x  x! ]3 b/ o8 C8 u& x
31. + W* I, e7 k$ l
    
32. %仿真结果与原数据对比测试 5 H' i: U1 S8 Z\" m3 t4 H5 {
    
33. x=1990:2009;
    ) v/ L3 j5 `% H( ^\" r
34. newk=a(1,:);, e' E2 ?( P8 J+ N3 c
    
35. newh=a(2,:);* s6 E0 F1 u( `
    
36. figure(2);4 S; }  f1 M( n7 q8 p( s! j
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');! L$ ]% h- b' Z9 [* O- f
    
38. legend('网络输出客运量','实际客运量');- o' `  a  C; _
    
39. xlabel('年份');ylabel('货运量、万人');
    9 A8 p5 M* W+ B! t\" Q& H
40. title('运用工具箱客运量学习和测试对比图');$ |; I: c7 H1 n9 Y
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    8 k7 T2 y( u  O% h7 l0 t7 f
42. legend('网络输出货运量','实际货运量');) Z7 L3 l( E& V5 |9 M. X9 u
    
43. xlabel('年份');ylabel('货运量、万吨');' I* X* v: m: Z8 C  @# w
    
44. title('运用工具箱货运量学习和测试对比图');, ^3 T- |6 a( D, W- }! M
    
45. 1 d# {6 y$ B\" c\" ^
    
46. %新数据仿真
    8 E* D7 B8 `: ~1 D1 W: a8 v5 Y( {\" U7 k
47. pnew=[73.39,75.55
    7 U\" q, k1 R7 t: m& O0 l) n/ Z
48.      3.9635,4.0975
    . l8 m& Q1 I( S4 h
49.      0.9880,1.0268];
    7 f( J- K) n5 {4 E8 t1 O, ^* U: `
50. pnewn=mapminmax('apply',pnew,ps1);
    ( u! f3 T1 m\" w
51. anewn=sim(net,pnewn);
    / @9 I. y. V2 e+ F+ V- K
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: g& [+ n- o, K2 ~+ p
   
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,...* X) Z, i2 X( X9 y9 O0 ?/ f
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   $ c7 }; w0 ?; W4 _  |& \* 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,...' }; p) K) K* p
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];. c6 m9 f9 h8 Y2 H3 t\" R( T
   
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,...( r5 o) S1 ~6 P' r) |
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   $ A: |# V% b; o- f( X5 Y
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...% E' P- e2 j6 ~3 p5 @! `* x
   
9.      22598,25107,33442,36836,40548,42927,43462];
   ) t: E8 h5 ]6 t# D4 {7 _
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...\" Y% w2 c7 N1 x+ [1 @# n' m
    
11.      13320,16762,18673,20724,20803,21804];
    - N5 }7 i' m) f0 ^+ n! q, m1 Y6 W
12. p=[sqrs;sqjdcs;sqglmj];) p: X$ k! E; J0 o# H; {+ {\" V
    
13. t=[glkyl;glhyl];8 M$ r0 f2 P! n0 k5 u; j$ x
    
14. %数据归一化/ x/ l, H  l' c& S
    
15. [pn,ps1]=mapminmax(p);
    ; ]  q! g5 s  \+ c3 S! k
16. [tn,ps2]=mapminmax(t);# X) v$ O3 m( C8 Z1 D
    
17. %dx=[-1,1;-1,1;-1,1];
    3 g7 ]. [  C- Q; U6 r. Q6 e( @) i
18. p0=minmax(pn);t0=minmax(tn);
    ) @* H1 H, l! |: u6 h
19. 
    , \4 s\" k\" v, |! M
20. %BP网络训练2 Q+ a7 K9 w. Y/ Z8 C
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); ! v/ D  ~; n4 A0 i3 c
    
22. net.trainParam.show=1000;
    8 q\" n\" Z. M* q0 F/ ^: _+ s
23. net.trainParam.Lr=0.05;8 W9 s0 Y4 u7 E: S$ x, K6 ?
    
24. net.trainParam.epochs=50000;
      ^  y+ F2 F, L4 r\" Q- s! j
25. net.trainParam.goal=0.65*10^(-5);, s8 |; m- K& E1 w( u
    
26. net=train(net,pn,tn);% F* e1 g; }! n& J0 e4 h' ?
    
27. 
    2 F\" P- M* _1 F+ ]
28. %利用原数据对BP网络仿真
      D! j0 y: V- J3 j  @9 G, f. }# A
29. an=sim(net,pn);
    5 B4 q. U3 q, @, V# ], E: m: d
30. a=mapminmax('reverse',an,ps2);0 i& u8 U! M7 W! N6 p  Y& _2 t+ e
    
31. : v5 L  P9 E8 ]0 _
    
32. %仿真结果与原数据对比测试
    ' w/ q. {0 X: z) L: @0 s' a) t
33. x=1990:2009;
    5 [2 I, t5 `! c( A, ]
34. newk=a(1,:);
      ~+ L9 J2 a, }; P0 u5 p: `6 y
35. newh=a(2,:);
    / a0 ?5 g3 ]6 o/ B  V: `; W
36. figure(2);
    8 q, S& |1 E8 [
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    ) M2 O( ~4 J) k* f5 m
38. legend('网络输出客运量','实际客运量');5 ^6 S: H0 c. s. C
    
39. xlabel('年份');ylabel('货运量、万人');
    1 F) U- `1 S3 Z* O8 `
40. title('运用工具箱客运量学习和测试对比图');9 q\" [/ h/ e3 \9 b\" W
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');2 h, ~) f\" o* r( Y
    
42. legend('网络输出货运量','实际货运量');: l- o; F8 j) r+ C+ W* u! F
    
43. xlabel('年份');ylabel('货运量、万吨');
    ; ~6 T0 D: x; S! z' V
44. title('运用工具箱货运量学习和测试对比图');1 S9 Y( s/ [0 x2 ~/ @! i
    
45. 
    . F3 ]  H! U, o! e# d
46. %新数据仿真
    1 }& Q' o& A' j4 Y+ ^) d
47. pnew=[73.39,75.553 J$ V  S& K+ j4 }: n
    
48.      3.9635,4.09753 S* ?( \2 L5 Y3 v' ?; o( j
    
49.      0.9880,1.0268];
      x# {: S0 \6 W. n% M! |7 @
50. pnewn=mapminmax('apply',pnew,ps1);
    ) e7 Z/ p* U: v' m2 k
51. anewn=sim(net,pnewn);3 Z3 u, C4 V0 D( z
    
52. anew=mapminmax('reverse',anewn,ps2)

复制代码

T-Eric

且生 发表于 2014-7-21 18:22
9 W, b2 @; C9 n/ g* w: a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3 h\" V9 @. f5 q
   
2. net=newff(p0,t0,6,{'tansig'},'traingd');
   1 [: T6 d9 A, _2 ]7 x/ _- D- @
3. net.trainParam.show=1000;# x# L. n% a& j  r7 Q4 v5 i: n
   
4. net.trainParam.Lr=0.05;7 p1 z% F% \  Y; u) N  y( }% W9 N/ m
   
5. net.trainParam.epochs=2000;
   4 N2 M. B1 A2 Q( ?1 X
6. net.trainParam.goal=0.65*10^(-4);$ \! z0 Y% s8 J9 d  [9 v
   
7. net=train(net,pn,tn);

复制代码

我只改了这里面的，
. z; i% |+ h4 r/ k: T7 c  {$ r: c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
三、迭代次数改成2000
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