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！！
- Y: V/ }% L0 y, H: d% }: r/ {
%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,...
6 p' L+ G9 \/ L) m7 L/ b6 t$ N1 f    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
1 c' ?7 ?! W6 I$ dsqjdcs=[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];
8 U( n0 s$ k; X" eglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
5 y0 I2 @, t: O" l' }6 L    22598,25107,33442,36836,40548,42927,43462];
1 W* O$ c: w! G7 e" o: Rglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
. d  u0 E* X- Q5 l7 ~, q    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

, p+ {* p; {+ H1 |" r. P0 C3 }0 c- s  ?%数据归一化
[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
dx=[-1,1;-1,1;-1,1];
& V) S6 ~1 f9 z; c/ J! m
$ r4 C; }/ i+ s%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
; E& j, r$ m0 V; V5 v, ?- `1 vnet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
9 ~2 V" b) g% L8 Y2 N2 c8 k1 E
* D: n, t1 h7 w  K7 ?6 ~- I3 g%利用原数据对BP网络仿真
an=sim(net,pn);
a=mapminmax('reverse',an,ps2);

6 _, V+ o5 @7 U4 A7 x1 ^* U%仿真结果与原数据对比测试
6 L  G2 L( ^9 K& d* Xx=1990:2009;
newk=a(1,:);
8 `) k8 i0 h* e8 Nnewh=a(2,:);
7 z! ^. ~, [$ ]& h0 q  j2 Ofigure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
6 v+ |( m( u# V7 a: |8 G0 |legend('网络输出客运量','实际客运量');
xlabel('年份');ylabel('货运量、万人');
3 W" Z3 ]9 j* E/ T; Btitle('运用工具箱客运量学习和测试对比图');
" T; U5 c3 c. I) Q( B/ W# h. X9 Bsubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
legend('网络输出货运量','实际货运量');
& C( V, e. I9 |1 p: l5 ?/ J; txlabel('年份');ylabel('货运量、万吨');
3 w- ~2 B5 l: \! ]. Ztitle('运用工具箱货运量学习和测试对比图');

  m  c, e8 S) a3 \' q( p/ I%新数据仿真
/ T' D8 t) C/ U6 d2 \pnew=[73.39,75.55
+ L6 |" P2 |, s    3.9635,4.0975
. v% n7 o% [; V( t- L    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
$ n7 z4 U* K. Y  h  c

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
. p+ N/ |1 H/ E* ?* C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 ...

' g5 r2 k! u+ t# N+ o3 Z- |因为版本更新，Matlab中的Newff命令用法有所改变，原命令为：
6 o8 i( @# ]- T# Xnet=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：
3 `  `+ a0 \/ X8 j/ w/ m2 p- e! o[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
a=postmnmx(an,mint,maxt);
pnewn=tramnmx(pnew,minp,maxp);
anew=postmnmx(anewn,mint,maxt)
& S; J: W9 y; I! \3 q我修改为：
[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
9 s' p% [; g5 `2 u) N. ]a=mapminmax('reverse',an,ps2);
pnewn=mapminmax('apply',pnew,ps1);
: t0 }' G# d) C( w. Ranew=mapminmax('reverse',anewn,ps2)
3 h/ o4 E7 r3 n
原程序为：
0 U8 |& e) |& O- @8 j/ t! `$ D0 Y%原始数据输入
clc
( Z& G8 D* J5 m- a9 j1 [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,...
- y9 W0 |  A2 {    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
7 X, m" s# o8 x$ S8 D8 d8 Qsqjdcs=[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 A+ N- n' c% H% Y" Y! q    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 G, A: w' W4 ?* o# K( }glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
* p; a5 b% W/ V3 J    22598,25107,33442,36836,40548,42927,43462];
0 o* P& S: t3 ]+ W: @/ Kglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
' z/ @3 s. r3 k1 Q+ U% z  n( I9 d/ d0 \    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
  X* O& D0 R5 n* t* e, {% |; L  pt=[glkyl;glhyl];
  p$ |3 Q2 _, Y+ `( |1 L" \3 M
%数据归一化
[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
dx=[-1,1;-1,1;-1,1];
% v+ i7 p& ^$ i$ X5 m7 Y
4 o4 y5 M# l4 v: f$ W6 A% u% h%BP网络训练
4 m1 ?( y5 d% E; m( Hnet=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
' o+ C" g* X# Y- w4 S: H% h7 c' Knet.trainParam.show=1000;
net.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
+ m. z* y# Z8 `$ ]net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
- |) h! B4 K0 `3 G1 b; U
4 S4 u! V# ?+ b/ P/ H%利用原数据对BP网络仿真
  }: ~+ o8 B$ n1 e* y  D0 Jan=sim(net,pn);
$ X" T# F# I: H/ Ia=postmnmx(an,mint,maxt);
$ S: m& X0 S( X, ]
  V; a4 n& B( g. L: O%仿真结果与原数据对比测试
1 U' ?) c7 e$ J# K7 ~" q* cx=1990:2009;
newk=a(1,;
, l' ]+ T, x5 v& b' @newh=a(2,;
( l0 I  ~+ ]5 h, N) ^! ufigure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
0 Y  D" }3 \* }. o) Ulegend('网络输出客运量','实际客运量');
xlabel('年份');ylabel('货运量、万人');
. T3 z* |! E1 f8 D# V- A# @title('运用工具箱客运量学习和测试对比图');
: G& j/ `. V. v( X6 C3 V, ~6 Lsubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
1 ^1 V1 {: B6 p2 w+ ?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');
5 j( U+ R- L  Q2 `
%新数据仿真
pnew=[73.39,75.55
    3.9635,4.0975
    0.9880,1.0268];
pnewn=tramnmx(pnew,minp,maxp);
anewn=sim(net,pnewn);
anew=postmnmx(anewn,mint,maxt)
3 c( @% y. H4 w" F; b. e+ j
修改后程序为：
%原始数据输入
* @, }% }6 f+ P4 z( S* B% J' i9 Bclc
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,...
1 I4 J$ v& B/ @" G" `    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
. I% F9 O5 x( ], qsqjdcs=[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,...
1 r; k% @  f: P% z/ `! o    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];
5 l% N+ ^2 Q; t. c) M. p: {' [glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
* s4 |2 a/ q4 D( i    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];
+ \6 ]  S/ w2 k' @; k* }$ n
( e2 J, g" o% c( G9 s, i%数据归一化
/ z% O  f7 O+ C2 c[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);

: A8 U$ o! D) K9 v8 n* j6 N%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
( N7 T5 ]& Z6 W  C$ Qnet.trainParam.Lr=0.05;
( v# ?* \  u* `4 b5 Cnet.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
; w* B8 K6 o4 d+ w1 Snet=train(net,pn,tn);

%利用原数据对BP网络仿真
- e/ v8 _1 n, van=sim(net,pn);
7 P) z( r' k& I; o4 B" C) c8 ca=mapminmax('reverse',an,ps2);

%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,;
9 i1 P  \' X2 ^2 L/ i7 N& Unewh=a(2,;
, P& u; I' n( Mfigure(2);
9 o# z$ [5 i" w2 b$ F5 P) gsubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
legend('网络输出客运量','实际客运量');
# F/ d4 ?! R6 E+ m( M& |  u& f1 vxlabel('年份');ylabel('货运量、万人');
& g' V! T, [! L& d8 y3 Mtitle('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');
2 h8 n. Q  j$ `* y+ y, Ttitle('运用工具箱货运量学习和测试对比图');

: l' G0 J; m; O8 c' w%新数据仿真
pnew=[73.39,75.55
" o# U. u  O4 O    3.9635,4.0975
    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
6 d+ Q1 }( q) {: h（修改的地方用颜色标记了）
. z) k( z- B' a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
   8 H( ~9 `9 u% `. A3 |
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,...+ H; |) T# i6 Z# b% J1 p3 ]
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   ; Y0 b+ j/ {  A- S9 w
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,...& C7 \5 ]- `! z5 \\" |. G
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];/ y  L! O+ n4 P+ e7 a# 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,...1 ~, h* E/ g8 o0 {/ v( \  z
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   \" i, _6 ]/ |% |* P
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   4 D0 B& @; t' F. o
9.      22598,25107,33442,36836,40548,42927,43462];
   7 i$ D: R2 P\" n\" J; ~/ ?
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...4 E& I0 v3 s. L, y# l, L/ I
    
11.      13320,16762,18673,20724,20803,21804];0 C# z8 i; `1 P- a
    
12. p=[sqrs;sqjdcs;sqglmj];
    2 y: @& ]# o+ O\" U
13. t=[glkyl;glhyl];
    % t/ c! K6 C; @
14. %数据归一化
    ( J% e) `( O& _$ s1 N
15. [pn,ps1]=mapminmax(p);
    % }  j- h/ k$ K9 p' l5 d% {
16. [tn,ps2]=mapminmax(t);5 F& q5 G) }0 y  {; O
    
17. %dx=[-1,1;-1,1;-1,1];
    7 o* Z\" W8 Q* a- H9 ?( f
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]# }4 u* D\" C. v* p
    
19. 3 Y& {$ R1 X0 B
    
20. %BP网络训练3 f, X% `+ U# z* B
    
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]1 U7 h. C/ C& l0 ]- u
    
22. net.trainParam.show=1000;
    * K7 ]  e! v$ E
23. net.trainParam.Lr=0.05;
    3 `1 ^: y/ G\" f% G# J, O$ B
24. net.trainParam.epochs=50000;
    6 H- Q$ B% e1 ?$ S7 E# I\" k1 N
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]
    / W- X& c2 P$ a9 ?
26. net=train(net,pn,tn);
    ) E\" U. h( h5 A% O\" v% P+ B9 |
27. 
    6 b3 }\" k& V6 G$ N9 k/ a. h
28. %利用原数据对BP网络仿真! A- m6 Y: a- c+ F; v
    
29. an=sim(net,pn);5 L5 x/ f' H0 e% X8 p( r
    
30. a=mapminmax('reverse',an,ps2);* ~$ y& ?) _: A% ^$ n  \6 E
    
31. 
    : S6 W. S( g  N. P
32. %仿真结果与原数据对比测试 % m) P$ f7 Q\" D
    
33. x=1990:2009;5 W: I, t4 [' ?% h
    
34. newk=a(1,:);
    ; V3 C2 h\" D  k9 x5 A6 j
35. newh=a(2,:);
    \" a* q8 @9 c! l( E: ?) [! D
36. figure(2);
    2 c6 p, k( U3 l3 q% P( `0 ^* I
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    - r0 f8 v' _+ A; O% R0 t1 `
38. legend('网络输出客运量','实际客运量');3 u  M) p3 x6 {9 u/ d
    
39. xlabel('年份');ylabel('货运量、万人');
    ' I' P) M7 @0 I4 J, h5 N
40. title('运用工具箱客运量学习和测试对比图');
    8 a' g2 b1 }# |$ z* g
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    $ @5 ~3 f! ^8 M7 b6 `
42. legend('网络输出货运量','实际货运量');
    & h- s# u! |+ t
43. xlabel('年份');ylabel('货运量、万吨');
    ! d( R0 ?$ m1 }/ Z- {6 t$ o. j
44. title('运用工具箱货运量学习和测试对比图');
    + U( p' Z8 H# V! I% e\" {0 [
45. 
    ' ~' Y  I; O1 m\" N8 w4 |
46. %新数据仿真\" w' c2 u5 F  Z6 N
    
47. pnew=[73.39,75.55
    8 _5 n* z* Y4 i, }
48.      3.9635,4.0975
    3 f: e: N2 r\" _* K' `# L
49.      0.9880,1.0268];0 t$ j1 f; ~9 v2 i; D$ Y- v# `6 y
    
50. pnewn=mapminmax('apply',pnew,ps1);
    ( c% I5 {1 j1 l/ _1 f1 d/ m
51. anewn=sim(net,pnewn);
    & l9 r% W- y3 \. a3 L
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/ V: S. {! {7 B. P9 g1 Y. 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,...& O) P1 K( W' O: Y) G3 G- @
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];/ Z7 c/ t* m0 G- Y3 a
   
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,...
   # ]9 F* ?+ m1 |. U2 c2 z7 \
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   ! p6 ]% P( O. N) @9 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,...$ s6 ^  ?! z0 d( P
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   $ C6 ]/ C' A2 M+ K, ?  a# {
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   : V. `+ i' E+ b3 e0 }7 ^/ b( a3 r
9.      22598,25107,33442,36836,40548,42927,43462];
   3 z' \' }8 O- E2 N
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    4 z) ~& _6 J% ^& J( }1 L; D, L
11.      13320,16762,18673,20724,20803,21804];; p; j: a- X8 d8 g
    
12. p=[sqrs;sqjdcs;sqglmj];
    ! A1 R/ ^4 A. _0 s' i, s3 N
13. t=[glkyl;glhyl];) w1 Y9 p( ~& q2 ~# S( e: H$ @  i) x
    
14. %数据归一化
    $ }% ?- ^7 L. e: e
15. [pn,ps1]=mapminmax(p);
    7 p* D( r( T- x  @6 P) P' H5 a, X
16. [tn,ps2]=mapminmax(t);
      ^2 Q3 E/ h! u' d, e4 J\" p
17. %dx=[-1,1;-1,1;-1,1];6 m/ i+ N$ w0 i' h
    
18. p0=minmax(pn);t0=minmax(tn);) O- J3 P& G9 H7 ?7 Y, Q
    
19. 
    ! ?' }* c, I: S
20. %BP网络训练1 B& B0 J$ R% a* m5 e: v; R
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); ( C. }5 B' Y\" F
    
22. net.trainParam.show=1000;$ o/ c4 q$ w1 Y3 d. I
    
23. net.trainParam.Lr=0.05;
    ( J0 M; f7 ^* E- J; v# v8 {7 g
24. net.trainParam.epochs=50000;, C5 L  u* R1 t: @6 _, N5 A
    
25. net.trainParam.goal=0.65*10^(-5);
      }8 r- t6 H4 s& D* V6 C
26. net=train(net,pn,tn);' \9 o; _: V- G/ `8 Z
    
27. 
    % T: V% j9 ]  r3 H$ B/ l9 B
28. %利用原数据对BP网络仿真
    8 o+ V$ s- Q: ?' }& `5 }3 u) i7 r6 b
29. an=sim(net,pn);
    . g; y% L- W3 Z$ p/ Q8 X9 v; G2 s
30. a=mapminmax('reverse',an,ps2);2 K# ?& ]; j$ q* p
    
31.   {! M3 W0 J8 V- \4 t& b+ a
    
32. %仿真结果与原数据对比测试 # e0 K0 l5 a, \6 g$ q
    
33. x=1990:2009;6 h$ p+ E\" x4 p2 E4 J* K; _, k+ @6 K8 W
    
34. newk=a(1,:);3 c1 s# i$ W5 B/ r2 }1 }
    
35. newh=a(2,:);
    \" T0 Z\" s# ~) I
36. figure(2);; Y\" l! Q/ {0 p; N4 w; b, l1 n
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    ' X8 V, n! n) x% \; C
38. legend('网络输出客运量','实际客运量');
    + N4 X  x* v/ S' F$ I8 ~
39. xlabel('年份');ylabel('货运量、万人');
    & l  u6 D6 x! A/ G! g) P+ d. A
40. title('运用工具箱客运量学习和测试对比图');# m\" N4 X- f; l) q: _/ Z$ ^
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    ) C8 \4 s& p. K) P1 c
42. legend('网络输出货运量','实际货运量');
    8 X) g8 x7 G+ q
43. xlabel('年份');ylabel('货运量、万吨');4 Q4 l0 `9 k6 x1 ?9 s
    
44. title('运用工具箱货运量学习和测试对比图');
    5 S5 T; k( F  i6 H7 `
45. 
    9 d7 ]3 L( ~+ t% h
46. %新数据仿真
    ' k2 K5 g6 _# X* i' A
47. pnew=[73.39,75.55
    # J4 l) g7 D! P
48.      3.9635,4.0975
    0 d- v8 K8 G2 Z8 n/ b  l: I\" P
49.      0.9880,1.0268];! n' [( d3 o; p' v8 B
    
50. pnewn=mapminmax('apply',pnew,ps1);
    , P\" A/ u; B9 D+ ]. J# U, s
51. anewn=sim(net,pnewn);( c2 J( d3 N' s
    
52. anew=mapminmax('reverse',anewn,ps2)

复制代码

T-Eric

且生 发表于 2014-7-21 18:22
- a- @+ k7 N6 ?8 Z. l3 o忽略我上面那个，改动的地方在18，21，25行……求批

* d  e6 T' Y' l: @2 C% ~, R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
   - l) I1 ^7 F, o/ q( Y6 y
2. net=newff(p0,t0,6,{'tansig'},'traingd'); - O  ?' q5 I\" i* Z. ~7 U
   
3. net.trainParam.show=1000;
   7 b+ B  o- u/ f$ S' u9 S2 {8 t
4. net.trainParam.Lr=0.05;9 o0 o! |6 w! ~8 c. f7 H% B
   
5. net.trainParam.epochs=2000;
   2 k# \4 F. y  ~\" X. A
6. net.trainParam.goal=0.65*10^(-4);* L3 o0 B6 R: X2 Y( 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
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