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！！
3 k. P# ?; R6 l9 l
" ~/ K5 V6 b' _& \7 d. J%原始数据输入
+ x* |: W5 L( ~6 R* P- o* B' jclc
* h9 p: Z6 _1 c0 s6 y$ W' wsqrs=[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];
" p- Q2 V6 S& Fsqjdcs=[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,...
7 e( }9 W& w7 k& Y$ Q/ w' Q  e    0.56,0.59,0.59,0.67,0.69,0.79];
+ N5 h. S6 L3 t$ H8 m$ Q9 U9 qglkyl=[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];
p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

%数据归一化
4 X; B" w+ s8 @# B  |[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);
dx=[-1,1;-1,1;-1,1];
5 N# {- V0 i, a' P$ \! l
%BP网络训练
& I; t8 a- E( ^: L2 |7 t2 xnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
# U# E6 @' S6 W; K4 Knet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
8 a+ G) _9 x' C) n* `+ ?net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
! ~' e+ [) R" L6 M* U: ^
* t. }$ l6 U9 `4 B" P%利用原数据对BP网络仿真
5 ?* }0 d9 d* @an=sim(net,pn);
4 T5 O# L( s8 i' Pa=mapminmax('reverse',an,ps2);
! G. J1 n& P" y4 _* }/ b
" O8 |) e7 [3 P0 k%仿真结果与原数据对比测试
x=1990:2009;
/ W' K# w, _  ^* S( znewk=a(1,:);
) ~: K3 l$ X. O/ Inewh=a(2,:);
figure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
! D7 i& U0 o( j$ |& T1 Xlegend('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');
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');
3 o7 s, M" O# `/ a: i  R3 _$ ~title('运用工具箱货运量学习和测试对比图');

5 U% z" G2 A0 D+ a%新数据仿真
pnew=[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)

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
6 v" }; U+ {5 f) h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 ...

: C" I' l8 v  M因为版本更新，Matlab中的Newff命令用法有所改变，原命令为：
# W; \9 q# @7 w. |. hnet=newff(dx,[3,7,2]{'tansig','tansig','purelin'},'traingdx');
我修改为：
9 m* H+ [' o* l; d9 L( k0 d0 enet=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);
a=postmnmx(an,mint,maxt);
8 m6 W; A) g& _) _4 G/ q5 Ipnewn=tramnmx(pnew,minp,maxp);
6 @& u6 ]& x- P. Canew=postmnmx(anewn,mint,maxt)
: p5 _8 ?+ G$ J4 ~9 ~6 K, n我修改为：
( n2 \- S& L8 G- B, M2 T[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
2 S; }$ M/ I& e5 S3 ta=mapminmax('reverse',an,ps2);
+ U2 ^) M3 H! ]: b' C2 ^" X- @, m0 Npnewn=mapminmax('apply',pnew,ps1);
- P: e8 r, {8 I/ V8 Aanew=mapminmax('reverse',anewn,ps2)
4 G1 D$ z9 S, Z/ g3 X& A2 x
原程序为：
%原始数据输入
$ Q9 m2 R" O  a, j8 ?clc
  |% D0 Q  Q2 ]9 `0 T7 X$ e+ n$ r% Rsqrs=[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,...
" Q  X' L" |0 y+ S- ]    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,...
( S4 C% c/ x9 _( k6 f    0.56,0.59,0.59,0.67,0.69,0.79];
; p. K3 c8 V3 v4 ]* ~- ^glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
) t. J8 P4 Y8 K' z# K    22598,25107,33442,36836,40548,42927,43462];
! n. r8 t$ E+ K: a: C, Eglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
! b5 N5 t! a; ?' m! H% x% t1 rt=[glkyl;glhyl];
0 R' w1 B' C" K3 G+ g! F
%数据归一化
, U/ P/ [" W2 v[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
* O; h( \& N& w0 T% c9 a! K$ ndx=[-1,1;-1,1;-1,1];

$ {* J2 U( ?, k, S. |/ F: Q%BP网络训练
# ^$ G: Q9 i) y) u, z  `net=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
& Y" n& L( P3 k6 x" X! V7 Y0 O' F' Knet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
. J+ Z3 N. Y- rnet.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
5 L- c% b7 d- I" E% \/ T
. m' ~; C$ M8 ?/ ?  o%利用原数据对BP网络仿真
an=sim(net,pn);
a=postmnmx(an,mint,maxt);

%仿真结果与原数据对比测试
x=1990:2009;
/ Z1 C" Z0 W. H6 \& Unewk=a(1,;
6 `; M4 b8 }6 e6 Lnewh=a(2,;
figure(2);
% l5 D& _9 Y5 U6 Xsubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
4 A9 ]8 g" g- F/ Q6 ~& G  }legend('网络输出客运量','实际客运量');
xlabel('年份');ylabel('货运量、万人');
2 j1 P8 i7 a6 R" J: Gtitle('运用工具箱客运量学习和测试对比图');
3 L5 F9 _* Q0 d% B" Msubplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
/ h0 O5 [6 e7 t* alegend('网络输出货运量','实际货运量');
% O5 h8 J2 `$ F: u8 u& uxlabel('年份');ylabel('货运量、万吨');
2 K. M. @! l8 ?title('运用工具箱货运量学习和测试对比图');

8 H0 s1 C7 j7 L: C4 F%新数据仿真
pnew=[73.39,75.55
    3.9635,4.0975
    0.9880,1.0268];
pnewn=tramnmx(pnew,minp,maxp);
% ]& {% B8 U6 ]  |8 Tanewn=sim(net,pnewn);
anew=postmnmx(anewn,mint,maxt)

修改后程序为：
%原始数据输入
2 L# B/ }3 Y8 I9 f. F8 X1 [+ `$ Mclc
* U( m4 w4 i- 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,...
( l3 s7 p; j6 P; t    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];
! G# y# M" |- o% r, Msqglmj=[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/ a3 T: m0 w; c  n) K1 N8 U    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,...
& S, e, f" d3 D$ i8 E4 m: p    22598,25107,33442,36836,40548,42927,43462];
glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
7 u( h+ n' f. u5 g! S    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];

& P8 H0 e  d7 t# ^' G1 s%数据归一化
& i8 I4 I, q& q' L  G: S8 I[pn,ps1]=mapminmax(p);
, w- t) s1 f5 i) R3 S[tn,ps2]=mapminmax(t);

%BP网络训练
! z" N; [5 _; Wnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
7 \( Q# t. A4 Q0 Dnet.trainParam.Lr=0.05;
net.trainParam.epochs=50000;
- @5 ~- w5 _4 g, F0 [) `6 Xnet.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);

%利用原数据对BP网络仿真
  I8 J  Y( [. T" x. ^" X+ T. Aan=sim(net,pn);
! j% i- {# a9 Z8 ~; o2 g; ?a=mapminmax('reverse',an,ps2);

6 N5 j. C9 A- A+ U' _2 {%仿真结果与原数据对比测试
& z  W1 s7 P0 y' {2 m" fx=1990:2009;
newk=a(1,;
newh=a(2,;
figure(2);
. v3 g5 \; x/ J2 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');
8 z2 s6 V8 ~5 |/ R! ]1 D( j& sxlabel('年份');ylabel('货运量、万人');
" Q, a4 ]  j4 h9 k+ o4 @( Mtitle('运用工具箱客运量学习和测试对比图');
$ r) n5 I& i- l& }( t- U) A: msubplot(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');
" |0 l9 h4 d1 l2 }$ z+ t+ |) b4 _
6 c) ~" m9 K# ~7 m%新数据仿真
pnew=[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)
（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  s& Y' r3 _8 K6 n6 U4 G! X; N# `
   
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,...
   : Z& N/ E7 S5 {\" J7 G
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];8 P: S* n5 I2 B% J5 M8 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,...\" Z- r$ }$ C8 g' _
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];* |1 M, f, y* N* k/ 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,...
   ; Y, K\" w& v  J: m1 ]0 x& Y
7.     0.56,0.59,0.59,0.67,0.69,0.79];% \9 x4 j- f  h0 p1 z4 ]
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,..., l+ [* {; i0 q2 @5 f7 L
   
9.      22598,25107,33442,36836,40548,42927,43462];: w4 M, I) T: c! u) A
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...3 I\" l# `2 \3 ?4 @2 C, |6 S$ Q* C
    
11.      13320,16762,18673,20724,20803,21804];
    % ~% s% ]# e- Y) Q& {1 f* f
12. p=[sqrs;sqjdcs;sqglmj];+ Z4 Q( J* k- v4 k
    
13. t=[glkyl;glhyl];
    & s- @% e  M/ ^% j$ k4 ~9 q
14. %数据归一化\" d) ^& K+ d7 ]4 `0 d7 D4 t* R
    
15. [pn,ps1]=mapminmax(p);
    % q, A4 h; j8 O( \% ~, N! @
16. [tn,ps2]=mapminmax(t);
    6 K% ^# O% \8 q$ U6 ~) F, E& w  I; ~4 S
17. %dx=[-1,1;-1,1;-1,1];
    ; R  X5 U1 [: u# O
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]9 U& w; \- l$ R* ]. ]8 z& v
    
19.   L# w1 ^. j5 z& e
    
20. %BP网络训练
    6 ~- e& D6 _- |; A2 _/ x& C0 I
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]5 s# R/ Q5 n+ i. u
    
22. net.trainParam.show=1000;; q/ m1 g% z' E) l3 ]& L
    
23. net.trainParam.Lr=0.05;
    # l2 `) U) g1 W9 `/ b
24. net.trainParam.epochs=50000;! n+ E: p0 P# F1 @* L
    
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]- S$ u6 h4 s7 `! Y\" w
    
26. net=train(net,pn,tn);; X2 [, [) u+ L9 V2 G6 k2 b. w/ U
    
27. ) L\" p6 I! n- N' q: h/ o
    
28. %利用原数据对BP网络仿真7 `: m# j) z& w& L3 t7 b: I
    
29. an=sim(net,pn);
    4 c& h& Z\" D/ }
30. a=mapminmax('reverse',an,ps2);8 a& T( G9 j; R8 ?$ X5 K
    
31. 1 U9 p. C$ l- b+ t
    
32. %仿真结果与原数据对比测试 ( E' L* x' p! c\" B( r
    
33. x=1990:2009;
    . Y$ S) r) ^& Y) ~$ Z\" w: N
34. newk=a(1,:);
    / `6 t& u2 g+ H& p! g1 v1 d+ L
35. newh=a(2,:);7 N( }! \2 M7 C( u  a
    
36. figure(2);
    9 Y  _0 _$ W( ~; y$ h& L
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    * F: l2 F+ f/ J1 K, B. Y
38. legend('网络输出客运量','实际客运量');$ T+ Y' b/ M9 l; q4 k) Y5 ]7 w
    
39. xlabel('年份');ylabel('货运量、万人');- s: U& p\" Q% I' R$ H
    
40. title('运用工具箱客运量学习和测试对比图');, Q6 p$ ~9 x/ n\" }8 n7 C8 S
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    ! l# t9 W$ m& w8 d
42. legend('网络输出货运量','实际货运量');
    / B# P' g( ?5 M$ F9 t
43. xlabel('年份');ylabel('货运量、万吨');
    & l8 T1 s2 R8 E, j2 G4 F: t
44. title('运用工具箱货运量学习和测试对比图');
    - `  @! k- Y4 k4 N
45. ; ~- C9 m  O6 L/ j' j6 ^: d2 p
    
46. %新数据仿真% [- }0 [+ u4 ?\" n5 D& E
    
47. pnew=[73.39,75.55' u; I/ G, i: z
    
48.      3.9635,4.09757 p: S) |! n) E2 q0 l3 Y
    
49.      0.9880,1.0268];
    2 j/ r1 U% W\" M  H
50. pnewn=mapminmax('apply',pnew,ps1);
    * A# u' o; R5 ~, \) T3 l' d
51. anewn=sim(net,pnewn);) U3 V; [) n# F: [$ N  X2 m3 g4 X! 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
   7 P% K- V8 o  m
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,.... p0 ~$ u; ?, o! I
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   ! J9 Z' \0 U$ d\" ^5 |
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,...$ d1 U) k) c% o' S\" ?( L' T, [
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   5 ~9 I* \: l2 J
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$ R; v( y) N7 p* `
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   & l3 E/ ]; L' N! Z
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   2 D( U- b. z+ V  A
9.      22598,25107,33442,36836,40548,42927,43462];
   7 b8 |8 D# _. H3 S; G( Y
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...0 [9 M4 [: B  e% F  d# n0 z8 r( y* K
    
11.      13320,16762,18673,20724,20803,21804];
    9 e9 @7 i& w4 C1 k- R/ Y
12. p=[sqrs;sqjdcs;sqglmj];
    8 j7 E0 `$ V* k, O& Z6 S/ {0 T
13. t=[glkyl;glhyl];
    ' A1 I8 c* R/ P0 [* n$ s
14. %数据归一化
    : a+ H: a, `7 J6 {8 b
15. [pn,ps1]=mapminmax(p);\" V5 S) K: Q( |$ N6 S
    
16. [tn,ps2]=mapminmax(t);1 T5 b/ `6 F) {9 p% Z: g' U: A
    
17. %dx=[-1,1;-1,1;-1,1];
    # \7 S1 J( h8 @0 O
18. p0=minmax(pn);t0=minmax(tn);
    1 D- f3 o1 w0 `9 z3 m: I
19. 
    5 j# b9 A, h, O; a: D\" \+ t
20. %BP网络训练
    , M0 n8 L) X- k  N' G1 ^8 ^  j- \
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); 6 Q6 @2 B) }8 h
    
22. net.trainParam.show=1000;2 l* c. J9 p\" N0 k\" M, L! I2 V4 p
    
23. net.trainParam.Lr=0.05;
    ! r$ y. p- t  e: z! d
24. net.trainParam.epochs=50000;5 K% i* z+ }: ?\" l- `) y3 ~* K
    
25. net.trainParam.goal=0.65*10^(-5);* K# l2 E5 E3 ?
    
26. net=train(net,pn,tn);! J6 P' u) v9 o+ c/ B4 H& B
    
27. 
    2 m( e  l) g7 G, i; B
28. %利用原数据对BP网络仿真( a6 X: e3 @\" S, M\" \
    
29. an=sim(net,pn);\" l/ j& k) Y, }4 H- _7 P
    
30. a=mapminmax('reverse',an,ps2);0 e/ m. d  t+ y( y# o, H
    
31. 
    \" n1 }# Q$ B6 \) W6 m
32. %仿真结果与原数据对比测试
    / h! Q- b3 ?9 ^6 A
33. x=1990:2009;0 S5 j\" X\" f\" R% r
    
34. newk=a(1,:);
    0 c# ~\" L1 g  h4 i2 D* C
35. newh=a(2,:);
    4 x( E3 y+ B9 _3 M
36. figure(2);8 V* {8 w7 Y9 s9 ]6 d7 O
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    3 B. X\" r5 X( b7 c
38. legend('网络输出客运量','实际客运量');
    \" B  q5 v/ i' F1 J
39. xlabel('年份');ylabel('货运量、万人');
    8 @: B  _3 n7 x/ z
40. title('运用工具箱客运量学习和测试对比图');
    2 L6 I) P: m& E' R7 w6 u
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
      _& }: c! D% [) {% l
42. legend('网络输出货运量','实际货运量');
    ' J  b5 W- d% Z) X
43. xlabel('年份');ylabel('货运量、万吨');5 r$ w\" h7 y1 ?* e\" X# }9 K9 c
    
44. title('运用工具箱货运量学习和测试对比图');4 g( L3 Q% M2 T! d4 V: l
    
45.   g' C$ `( S. R& ^
    
46. %新数据仿真* o* c& Y* F. t! p1 r
    
47. pnew=[73.39,75.55
    / D\" B' Z  W2 s+ I4 L# `
48.      3.9635,4.0975
    6 v3 Z  h. T: _5 f5 o5 X: ~* c
49.      0.9880,1.0268];
    6 C1 {# N! C9 n  S
50. pnewn=mapminmax('apply',pnew,ps1);\" Z8 W3 t6 A! ?/ E7 e
    
51. anewn=sim(net,pnewn);
    7 A8 {, i# {2 \' a
52. anew=mapminmax('reverse',anewn,ps2)

复制代码

T-Eric

且生 发表于 2014-7-21 18:22
6 G2 J9 z& ?5 n: q$ Q% ?! }& }忽略我上面那个，改动的地方在18，21，25行……求批

7 d8 I4 X% k/ D& N- N8 q/ k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
   6 W, n: W7 O) @& ?/ H
2. net=newff(p0,t0,6,{'tansig'},'traingd'); , r! H5 g/ X( N9 v) p2 G
   
3. net.trainParam.show=1000;: Q9 u- e  n/ x: N  R
   
4. net.trainParam.Lr=0.05;
   ) Z7 ~6 d0 v2 i) _4 Y0 d% q
5. net.trainParam.epochs=2000;
   * K4 ?( A# n8 K2 R\" Y! J! ~
6. net.trainParam.goal=0.65*10^(-4);4 }0 \8 r/ a  U9 h6 f& y4 L
   
7. net=train(net,pn,tn);

复制代码

我只改了这里面的，
+ c2 t% }4 a2 t% e! k: `! ?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。
1 z9 g- \' ^' D! ~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