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！！
  ?* I& Y% W7 q" m3 `
& l/ `3 P3 n: {3 {  o%原始数据输入
/ N4 Q. v" F" C9 qclc
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,...
0 L4 v: A9 r+ S4 o5 S# Z% b9 c0 U5 t    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
7 ^; X9 ?# x- f, Psqjdcs=[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];
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,...
" B- g! V  l* F3 S* M3 C& q1 _    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
! _4 J6 A- I. ^4 [+ P+ q+ Rt=[glkyl;glhyl];

' m! a5 X; I* f% O5 C%数据归一化
& {: u) G! _2 }2 J. K' P- X[pn,ps1]=mapminmax(p);
1 S& ~5 r, f* n% L" F[tn,ps2]=mapminmax(t);
dx=[-1,1;-1,1;-1,1];

4 Y% @9 w" j- `1 F# ^%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
* P' l: G$ X4 B) j3 A; d7 Pnet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
% q3 v* v2 Q! r( [; v. L/ j: Vnet=train(net,pn,tn);

%利用原数据对BP网络仿真
5 K8 ^4 X/ c/ K0 D3 Z  van=sim(net,pn);
a=mapminmax('reverse',an,ps2);
# {9 P/ X: i8 o3 l& C% S' K( q
2 _$ _8 b. R/ W# N%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,:);
newh=a(2,:);
9 Z; m! k$ n% v3 E" ~! Xfigure(2);
% |% R: K2 ^  p. u7 ~subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
- \9 s* Q, N/ V* `) Ilegend('网络输出客运量','实际客运量');
xlabel('年份');ylabel('货运量、万人');
% U0 I$ i. M; L$ G) ititle('运用工具箱客运量学习和测试对比图');
7 M9 f5 Y' k/ h- g" j" G: Bsubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
5 u$ A, J4 E: @3 Z) Y! @; Jlegend('网络输出货运量','实际货运量');
/ z9 l# u6 j/ N: c+ s) E4 @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');

! L- J9 x3 n3 X/ S$ i+ A, b%新数据仿真
' m+ ~% p) K3 h; Hpnew=[73.39,75.55
    3.9635,4.0975
    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
7 ]+ w0 m: }1 z, L  C8 |& t# banewn=sim(net,pnewn);
9 {; M; S  Q0 U# \anew=mapminmax('reverse',anewn,ps2)

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：
+ D8 \0 r  ^# S3 f* e$ ~7 [net=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：
[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
: L  V: c$ z1 b7 }+ z# E- Y* E$ qa=postmnmx(an,mint,maxt);
( c7 S( A8 P2 a+ t% ^- E  r6 bpnewn=tramnmx(pnew,minp,maxp);
anew=postmnmx(anewn,mint,maxt)
, L- ~' B: F$ A( X1 G: ~我修改为：
8 u) u. V0 R4 o[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
a=mapminmax('reverse',an,ps2);
pnewn=mapminmax('apply',pnew,ps1);
! o8 F+ R1 V& T! manew=mapminmax('reverse',anewn,ps2)
7 z* s+ t6 Y6 i3 P, T. r
; V2 L# J% U7 m' c# H' V原程序为：
; U, X5 K) V2 T# Z1 a6 L- ^3 X) B%原始数据输入
clc
/ l3 a( ~$ p0 I! nsqrs=[20.55,22.44,25.73,27.13,29.45,30.1,30.96,34.06,36.42,38.09,39.13,39.99,...
% v8 }$ E. c# }9 \8 c, c' B+ A    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
$ @- k9 v9 W+ O9 usqjdcs=[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];
/ [# C# M8 P+ C- esqglmj=[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];
; E. X4 ]- A# wglkyl=[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,...
    13320,16762,18673,20724,20803,21804];
( H9 A" v: {. @3 yp=[sqrs;sqjdcs;sqglmj];
9 M/ I. c9 G$ n, et=[glkyl;glhyl];
: P) z& Q+ Z3 e" B$ c( ^8 [
%数据归一化
[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
dx=[-1,1;-1,1;-1,1];
; W# M# w/ ^* S' ?/ g+ Z
7 @# J9 z1 o5 Z0 x%BP网络训练
( F; s; H" ?# s- M1 Mnet=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
net.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
  M+ G/ ]3 n$ D$ g, p. Fnet=train(net,pn,tn);
9 G3 U* q$ D- M' i  i- m+ F5 p2 c
%利用原数据对BP网络仿真
0 r& n+ y, n; m# Z" M' t' Q  ean=sim(net,pn);
a=postmnmx(an,mint,maxt);
: u( V; r) N& j; N% C: P/ g
5 a/ d3 r3 m9 ]( c3 W! L$ l/ H%仿真结果与原数据对比测试
x=1990:2009;
$ Z' h* v* h( q2 G3 Hnewk=a(1,;
+ }% c" o; v+ P3 Qnewh=a(2,;
figure(2);
: X6 g, @9 W  {* g  v! z4 J* b/ L* Bsubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
! w! c& w" C7 I8 |2 j$ blegend('网络输出客运量','实际客运量');
xlabel('年份');ylabel('货运量、万人');
! h" v8 T+ _( e; Atitle('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--+');
8 t$ s) g! T& R9 G/ U% Q8 [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');
9 |6 P+ K- \5 B4 d/ o) _/ n
%新数据仿真
pnew=[73.39,75.55
+ V. o; A( T/ B$ c, c    3.9635,4.0975
* s+ H! p7 e$ R# I* x, O    0.9880,1.0268];
pnewn=tramnmx(pnew,minp,maxp);
$ c- x7 Q" P( F! |% Z2 eanewn=sim(net,pnewn);
$ `/ R) m' y+ H/ s2 I" W+ Ganew=postmnmx(anewn,mint,maxt)
3 X4 T9 ]" [( ?9 t7 g/ i
修改后程序为：
%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,...
: V( S) U' W  d/ V    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
- D, J) M1 b0 j# u$ ?! ?7 H9 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];
glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
    22598,25107,33442,36836,40548,42927,43462];
) o2 z7 c9 R/ c' o0 s8 q' Mglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
- R2 Y9 L6 u/ u3 u0 x    13320,16762,18673,20724,20803,21804];
, v0 }& H. ?9 F# o. L! }  Bp=[sqrs;sqjdcs;sqglmj];
. J" P  k5 I' s  c6 `t=[glkyl;glhyl];
+ g: a5 X+ _+ c( L. {6 v
  e- Q# @+ W" {5 C$ n6 V, {3 j8 Q- {%数据归一化
[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
% Q5 x- h8 I' z
%BP网络训练
$ I$ V2 o/ V) q: W$ H: bnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
( y( G3 U; ^" r3 gnet.trainParam.show=1000;
net.trainParam.Lr=0.05;
- X/ T, Y. V# F$ q, G1 G1 u7 ]3 Wnet.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
% g- V9 ~' q% S- a) ban=sim(net,pn);
a=mapminmax('reverse',an,ps2);
9 V5 O1 V0 c9 w( Z( k, h9 R$ J2 M
- _6 _+ I# n8 j%仿真结果与原数据对比测试
x=1990:2009;
' e. ~$ ?4 ]) x- l$ A2 cnewk=a(1,;
newh=a(2,;
6 x  d5 h5 H  y. i; `2 Hfigure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
8 q) V" r8 F. n5 B; \legend('网络输出客运量','实际客运量');
; R9 d1 @3 l: V3 Uxlabel('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');
( C9 z# ?: ~$ D& ]; m) K( lsubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
: l# k) E- W# n# hlegend('网络输出货运量','实际货运量');
0 X, D$ B9 G3 kxlabel('年份');ylabel('货运量、万吨');
4 f. Z+ i: P7 P/ Ttitle('运用工具箱货运量学习和测试对比图');

" F; C0 {- T4 C" r+ @; w1 Q3 T%新数据仿真
# N+ K; s+ X0 tpnew=[73.39,75.55
    3.9635,4.0975
    0.9880,1.0268];
1 ~( E: R! r$ jpnewn=mapminmax('apply',pnew,ps1);
) f: L. q# g( _+ a6 Banewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
) a! i2 D4 x# y2 F& T/ w9 I7 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
   3 ], D* t9 D( o. W8 z9 N3 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,...6 f' @( E$ T+ p  W
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   7 o1 v8 L+ g. x8 R8 f
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 a5 R& p% [
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   / i5 D) a/ F) y9 T, @8 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,...
   ' C% S+ D  n+ m9 T- w
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   $ c3 c4 O6 N; ?) z9 o: h! n8 w# x
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   + b7 `  c0 I. ]6 S% B' H& P: z; m( r9 B
9.      22598,25107,33442,36836,40548,42927,43462];  J0 U6 t' i4 K. v3 \& v& v8 E
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    8 Q7 ?\" i/ J; g3 M6 O  j1 U# F% t
11.      13320,16762,18673,20724,20803,21804];$ e$ }/ _+ u3 o5 Q' _+ z5 f8 `
    
12. p=[sqrs;sqjdcs;sqglmj];, q\" V& c8 i1 A5 s
    
13. t=[glkyl;glhyl];
    % H) m( B* {7 p$ j! L1 `+ t
14. %数据归一化
    3 P6 r$ w. [$ l8 Q! p9 \
15. [pn,ps1]=mapminmax(p);/ B  J: @\" e: [
    
16. [tn,ps2]=mapminmax(t);
    9 d' B8 a+ B) l4 s$ n
17. %dx=[-1,1;-1,1;-1,1];
    # O/ C9 g) M! H6 I0 l7 K* d
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]9 Y! Z4 \9 S/ J
    
19. 
    3 p2 A1 [( Y' ]  p3 X: y; n
20. %BP网络训练
    ' U% P) C0 b- O' m
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]  {; @( i. G5 ?( z2 W
    
22. net.trainParam.show=1000;
    + W/ q# D) d2 I3 {
23. net.trainParam.Lr=0.05;
    & T8 u3 d  ]( e6 W; T( @
24. net.trainParam.epochs=50000;; Q% t; A1 j8 B7 C2 x\" W8 b
    
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]
    8 F' K- D\" l; S8 G5 {& D  h
26. net=train(net,pn,tn);0 e( Q) \' A1 v) f) G
    
27. 
    ( v8 A2 Z0 Z: V+ H) V, t
28. %利用原数据对BP网络仿真
    ) s- ?9 [5 k2 p3 g2 @
29. an=sim(net,pn);7 O$ {0 c8 j* [+ g# j( T
    
30. a=mapminmax('reverse',an,ps2);
    $ k6 w( \. P+ _# X, Y  L$ n/ ], X
31. % ]! f& c9 K5 P8 m
    
32. %仿真结果与原数据对比测试 ( z\" h3 `8 e, V: E4 _
    
33. x=1990:2009;- ?0 H* v3 `1 G% a- g1 \
    
34. newk=a(1,:);& k& V: [& \: E9 ^. S5 e
    
35. newh=a(2,:);
    + L% o& e5 x# G) g' T
36. figure(2);
    7 L, q' X9 d  d3 I3 w! P
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    3 s4 B% J7 }0 h8 P% n( k. Y/ d
38. legend('网络输出客运量','实际客运量');& f8 L6 c/ D0 G
    
39. xlabel('年份');ylabel('货运量、万人');; p9 i; ^/ p- O& d# F
    
40. title('运用工具箱客运量学习和测试对比图');/ `- n. x! l9 _! P2 H
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');/ N9 a1 S) x! V5 d7 ^6 T, i
    
42. legend('网络输出货运量','实际货运量');
    3 G! e2 _( L9 u  E9 R& q
43. xlabel('年份');ylabel('货运量、万吨');/ w/ r1 Q% l7 t. U) G5 X
    
44. title('运用工具箱货运量学习和测试对比图');& S. y/ o# d6 L# M
    
45. / Z5 B! c$ V( ?, ~& L% o/ Y
    
46. %新数据仿真
    / U\" V5 S\" d0 [5 g2 h9 v/ f, v- t* N
47. pnew=[73.39,75.558 v9 z' H5 c& ?/ N$ K* q- n
    
48.      3.9635,4.0975' L) O& N3 R: M# ^
    
49.      0.9880,1.0268];7 i1 J2 R7 k! `& Q& S* M. D
    
50. pnewn=mapminmax('apply',pnew,ps1);
    4 B+ c\" v! H0 K3 T
51. anewn=sim(net,pnewn);4 a3 G\" W( g+ j+ I: f% p
    
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
   $ N# Y# j$ Y+ G
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) F; Q  \1 y$ \5 _# J9 |2 a9 ~, q: N
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
     b4 T' x, F2 @& i$ U
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,...\" O/ i; J: O. c0 ~/ v
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   ' s) w+ L  @! n: x8 T) a- U0 b
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,...
     m+ B* ~& l( o7 {0 [& P! w' M& ?- P5 v; q
7.     0.56,0.59,0.59,0.67,0.69,0.79];% e( a( }3 b0 _9 }
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   4 _- K2 }5 ]\" o9 j: \  L\" x
9.      22598,25107,33442,36836,40548,42927,43462];  p# d1 Q; w\" h/ W\" \
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...; v6 u5 N! s) D- p8 y7 i, y
    
11.      13320,16762,18673,20724,20803,21804];/ I! S) B0 [* i8 ]5 Y\" m+ [
    
12. p=[sqrs;sqjdcs;sqglmj];9 t9 ]0 `\" @% w/ V
    
13. t=[glkyl;glhyl];
    $ b* {2 D8 [0 {6 ?# \1 C- O$ m
14. %数据归一化; [+ q4 V& V/ v6 O3 @% |# ?
    
15. [pn,ps1]=mapminmax(p);, c$ K2 I* k; O* D1 V1 ?/ B
    
16. [tn,ps2]=mapminmax(t);
    ) l+ w& Q\" h( O1 {
17. %dx=[-1,1;-1,1;-1,1];# Y\" N* P: Q! ^1 {. S7 ^' B# G% O
    
18. p0=minmax(pn);t0=minmax(tn);
    . M5 M) ^3 d0 r+ t
19. 
    9 j$ a! L) U& a0 n% W) V  N
20. %BP网络训练, x+ v* j6 _1 e# o, c
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); % m; M1 ^! g% J7 v
    
22. net.trainParam.show=1000;
    / f3 D8 n! s! s9 P
23. net.trainParam.Lr=0.05;( q5 S6 Y# @\" w6 j# s
    
24. net.trainParam.epochs=50000;$ L5 Y+ W; M/ P' P; I/ O- h2 o, L* h
    
25. net.trainParam.goal=0.65*10^(-5);
    8 |# m1 k$ ~, E& g5 S
26. net=train(net,pn,tn);8 D1 ]' a1 C\" Q( C: |
    
27. : x/ A0 I7 D/ l9 q: s4 E$ ?\" S( u
    
28. %利用原数据对BP网络仿真
    1 ^1 M! y. z\" w1 {' `5 a9 N
29. an=sim(net,pn);
    - U/ ?: Y2 c5 t& t- R  @8 a1 D- f- s
30. a=mapminmax('reverse',an,ps2);2 B4 R1 R$ g; K/ \( I
    
31. : K+ l7 _, J$ N4 g& C4 |4 c
    
32. %仿真结果与原数据对比测试
    0 ?4 l0 L9 i' U, U9 j; l# i3 D
33. x=1990:2009;2 Y/ f2 e9 \! Y0 n* ?
    
34. newk=a(1,:);# A! j5 ?4 I3 ~6 Q
    
35. newh=a(2,:);
    # a6 V$ Z3 K9 `$ V
36. figure(2);
    6 R2 |) j& y1 z0 @5 I6 r
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    ; b' N9 f8 K, P+ N* s
38. legend('网络输出客运量','实际客运量');
    $ K7 s  \4 {- T) E- \' q0 g8 m
39. xlabel('年份');ylabel('货运量、万人');
    . N- L. g# ?) f5 Z5 t7 [
40. title('运用工具箱客运量学习和测试对比图');, t; Z# |# F; x$ V5 E9 Y
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    . |' p9 ]5 @, h. i
42. legend('网络输出货运量','实际货运量');
    8 j, d* k6 K7 G+ |
43. xlabel('年份');ylabel('货运量、万吨');
    \" D+ N+ x  [( j8 e( i. @! s; H
44. title('运用工具箱货运量学习和测试对比图');
    0 I* D9 G5 `& I/ f
45. 
    ) o8 o0 j9 `, I
46. %新数据仿真
    % d* {% g) o0 m, i: B) f' H
47. pnew=[73.39,75.55
    8 J! l4 T$ h+ a5 b7 k! ]3 I6 y
48.      3.9635,4.0975
    ( m( R& R& m; i. B
49.      0.9880,1.0268];7 b% q3 p% z3 p  y
    
50. pnewn=mapminmax('apply',pnew,ps1);1 w( ?' ]* E# ^' e1 G
    
51. anewn=sim(net,pnewn);
    \" C# P! {7 y: n! z( d5 N- B
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

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) S& F2 J' s+ I5 \, W( f
2. net=newff(p0,t0,6,{'tansig'},'traingd');
   3 R, Z\" o4 y  c: Q6 b
3. net.trainParam.show=1000;: p! @2 q! g0 [3 d
   
4. net.trainParam.Lr=0.05;# m$ R% O' U: T. a5 c
   
5. net.trainParam.epochs=2000;
     O  j, T# l. f; w+ H* {
6. net.trainParam.goal=0.65*10^(-4);+ I4 m7 x5 l( n8 w& {6 B\" ?8 _
   
7. net=train(net,pn,tn);

复制代码

我只改了这里面的，
' w' Y& ?5 V& J( j$ J% I$ r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
5 R/ G1 @; _( s3 {: l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。
. ]: t4 H" V% e1 i0 S5 A/ V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