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！！

%原始数据输入
# o9 V- ~- x# o. J! k/ F: Pclc
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,...
3 [3 q, i- h9 [  ^: o9 T$ h    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,...
, F  s. r0 ?5 _) `1 c. l* W    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,...
    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
2 u- Q) ~- C, R+ V5 |t=[glkyl;glhyl];
+ P, b) v  g% y& K
$ T/ Y7 S. n2 M0 M7 [; E%数据归一化
! |5 i/ x" Q( Z' J' R8 a[pn,ps1]=mapminmax(p);
, R/ l7 w- @8 N0 [3 e[tn,ps2]=mapminmax(t);
# B9 K- Q) ]. J5 Z2 [dx=[-1,1;-1,1;-1,1];
# e: I: B2 }2 e
%BP网络训练
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
1 P/ r* D3 a; Cnet.trainParam.show=1000;
; |8 R7 D- ~/ h* h* z  [- C0 Y3 Xnet.trainParam.Lr=0.05;
& Q) A. t, X7 d) B: fnet.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
5 r9 X9 G+ y! I* yan=sim(net,pn);
a=mapminmax('reverse',an,ps2);

%仿真结果与原数据对比测试
& M& Y  c* D  a0 }3 ex=1990:2009;
# s9 V# \9 A. @7 wnewk=a(1,:);
newh=a(2,:);
, C, Q+ ]) A6 M0 a6 p4 }figure(2);
0 Z" K$ y6 P3 Z3 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');
) }9 y6 c8 M+ d: \xlabel('年份');ylabel('货运量、万人');
9 I7 N, C* j+ r# f4 ]7 A* btitle('运用工具箱客运量学习和测试对比图');
# F) v, t( Z% Isubplot(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');
1 @% D& X# N) h8 @title('运用工具箱货运量学习和测试对比图');
+ z: B4 V* v( y3 Q
%新数据仿真
pnew=[73.39,75.55
$ R+ b* Y; D8 f# M    3.9635,4.0975
    0.9880,1.0268];
3 [! c( G$ X9 P$ Z: `8 Z9 cpnewn=mapminmax('apply',pnew,ps1);
3 V# {8 f8 k* ~  `: Q9 Wanewn=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
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 ...

6 l, k9 j( b* j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');
/ H( w, C/ u; o+ o/ K2 U# p我修改为：
9 B  M$ Z! }, ~' m( `1 ]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);
a=postmnmx(an,mint,maxt);
pnewn=tramnmx(pnew,minp,maxp);
anew=postmnmx(anewn,mint,maxt)
9 K. {! ^/ i$ L2 V6 S8 u我修改为：
$ i; m  n* K6 ^) j[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
a=mapminmax('reverse',an,ps2);
pnewn=mapminmax('apply',pnew,ps1);
7 d0 Q$ u8 v, T8 X& b! nanew=mapminmax('reverse',anewn,ps2)
: d9 M; ]! a% d- O$ K! W+ `
5 _4 j2 L1 p9 r' Y) ]0 ]原程序为：
%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,...
3 t& e" f1 O! i& {, o6 ~; P; w    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
* W% e" }8 I2 E0 W. csqjdcs=[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,...
7 Y4 f' B; v( D    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,...
$ ~6 a0 _$ I9 d2 L. D/ q% d* F+ ?    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,...
- T7 t9 C  A+ \    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];
7 q2 a4 W+ V% [
%数据归一化
[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
dx=[-1,1;-1,1;-1,1];

%BP网络训练
net=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
net.trainParam.Lr=0.05;
5 g* S4 H% u0 Q/ f9 unet.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
/ e+ M2 [+ s3 x: q; |- D. |net=train(net,pn,tn);
2 G# l) Y0 c( h% k9 T: B: B) `8 E. S
%利用原数据对BP网络仿真
an=sim(net,pn);
a=postmnmx(an,mint,maxt);

. l; `7 Z: r) c9 f: J/ F) t, e%仿真结果与原数据对比测试
x=1990:2009;
newk=a(1,;
# _8 u' ?/ `0 `* G/ Z/ rnewh=a(2,;
figure(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');
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');
9 y/ S! Y$ ?$ r, R' B& Q9 T$ ^title('运用工具箱货运量学习和测试对比图');
! j. r# G% j" H9 k5 L0 o
1 M3 ~  Z  O5 A: H%新数据仿真
) z+ B8 {# S; P: Dpnew=[73.39,75.55
    3.9635,4.0975
; d/ r5 J3 m* V" f* `, K2 R    0.9880,1.0268];
( [9 p  W: V% _7 E( Bpnewn=tramnmx(pnew,minp,maxp);
anewn=sim(net,pnewn);
anew=postmnmx(anewn,mint,maxt)
! n( ?5 M# i4 G1 g8 Q# d
1 j- u$ r4 |( q" p5 U修改后程序为：
4 W4 y/ \  F% i%原始数据输入
clc
6 [/ E# K. n0 `1 Lsqrs=[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 b4 v9 D6 e6 t2 m# G    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,...
9 F) A) ?4 \# P    2.5,2.6,2.7,2.85,2.95,3.10];
' ?: o' C( |/ U" \5 g- }3 w, P! Wsqglmj=[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,...
. R0 {  x. C' e0 I) P) {8 Q- F# V    22598,25107,33442,36836,40548,42927,43462];
" i* j6 R# `0 @$ i+ n9 zglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
. S* Q0 U1 P/ L  W, u2 F2 U    13320,16762,18673,20724,20803,21804];
. P0 q2 N/ A& K# l, g8 h( Dp=[sqrs;sqjdcs;sqglmj];
! K3 W8 @$ z* A3 e& ^+ W& [t=[glkyl;glhyl];
  v4 {5 b( l& u" `' ~+ ?5 c
%数据归一化
[pn,ps1]=mapminmax(p);
7 X, L' V) W+ G# i. L" Y[tn,ps2]=mapminmax(t);

%BP网络训练
0 b% C+ n% k8 ~) W& l9 ^$ Qnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
* k; J* M% d" r8 i* ^$ s1 V1 _+ wnet.trainParam.show=1000;
* `1 }) J1 R9 t# Nnet.trainParam.Lr=0.05;
net.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
7 S# {8 q3 U- a5 h5 Ian=sim(net,pn);
a=mapminmax('reverse',an,ps2);

%仿真结果与原数据对比测试
# X1 b8 @1 |9 rx=1990:2009;
+ n! U+ y4 |! q" R* D: Enewk=a(1,;
) p( `) s1 ]% |( Y1 t4 Onewh=a(2,;
figure(2);
  f) C/ w2 T, h, Nsubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
( H8 K1 V% w' g: slegend('网络输出客运量','实际客运量');
; F/ N$ R% ~9 R+ l2 j1 V( 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');
/ S- w" \7 W: l% r; c$ s9 ]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');
4 r" P( b$ P2 h, ~* Rxlabel('年份');ylabel('货运量、万吨');
; E; n2 m9 A6 l# E! }6 Utitle('运用工具箱货运量学习和测试对比图');

%新数据仿真
) Z9 p4 @+ W* J) J/ f0 k. N+ rpnew=[73.39,75.55
# X5 L( O2 s/ Z/ |    3.9635,4.0975
    0.9880,1.0268];
4 V" h4 w) B/ H/ z/ u, Zpnewn=mapminmax('apply',pnew,ps1);
4 u9 R# o8 z' S" h$ V( g' Oanewn=sim(net,pnewn);
$ j: c! b0 o( I' O9 ^) r/ [4 Xanew=mapminmax('reverse',anewn,ps2)
（修改的地方用颜色标记了）
% J. V2 x8 b$ z0 T' d* j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\" {/ \  w( e, J$ E& O\" X+ f; I
   
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,.... `  ?# k1 u) x3 T. N
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];' r7 Z4 Y1 p# J) c5 M
   
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,...
   ) e% {( U: x9 ^( S! d' z
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   ( B5 V' @( Z- Y) V8 c8 S
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,...
   ; i! c) a& `9 a+ h
7.     0.56,0.59,0.59,0.67,0.69,0.79];8 ~9 e8 p( @1 @1 s2 I5 }& U( U
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   : V) ~- J* p, i  i
9.      22598,25107,33442,36836,40548,42927,43462];
   + V# g' }7 S9 Q5 y; n. }) n6 G
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...2 Y4 g9 P! f2 s9 F1 ?' J- e
    
11.      13320,16762,18673,20724,20803,21804];+ q+ }+ ]% q2 ~, n
    
12. p=[sqrs;sqjdcs;sqglmj];, {; B; o  o\" F5 u/ }( B2 ^8 R
    
13. t=[glkyl;glhyl];1 A! j0 p4 k8 m( ]; O
    
14. %数据归一化
    0 Y' ]+ ~6 x7 U7 @. ?5 S) X- w
15. [pn,ps1]=mapminmax(p);
    ( y5 ^8 g( P  B' [# s\" |4 ]
16. [tn,ps2]=mapminmax(t);: t% D- }- g: R0 `) m
    
17. %dx=[-1,1;-1,1;-1,1];; ~4 s( I2 b4 z0 z
    
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]# [3 r8 U$ F1 U. f
    
19. 
    + d  X3 n7 D6 _; K, x( ^
20. %BP网络训练
    + ]; z+ v7 f! i  f7 U- {
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]
    ) L* }8 s/ _* K9 S2 X4 c
22. net.trainParam.show=1000;
    8 ]% e+ h: h3 Y/ d0 n+ l  r% w* N1 A
23. net.trainParam.Lr=0.05;9 x$ A0 \  S4 s
    
24. net.trainParam.epochs=50000;
    + f5 z8 k5 Y$ R3 {/ h
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]1 l+ H2 A( I9 j2 G& I$ d
    
26. net=train(net,pn,tn);
    5 P; S, z\" x  e+ ~+ X% B
27. 
    6 k5 B0 _( w0 @: X$ \( G; C) d9 x
28. %利用原数据对BP网络仿真
    4 {5 k+ w, Y; ~8 R2 s$ q. ^! i
29. an=sim(net,pn);  n' Y5 p: j5 x
    
30. a=mapminmax('reverse',an,ps2);
    ! ]0 s- g+ B2 o0 N
31. 0 s: R# u1 W. f0 G% m
    
32. %仿真结果与原数据对比测试 * X9 f4 C; E+ p2 I! r
    
33. x=1990:2009;* `7 ~! x; ]; R( s! b  M& z
    
34. newk=a(1,:);
      |; c( h( I4 R0 J+ p$ m  S
35. newh=a(2,:);
    9 C7 a$ p6 y5 u* A
36. figure(2);
    ; K+ }6 j# N# I0 K. g
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    3 p5 N5 B0 u1 N$ ]' E\" g
38. legend('网络输出客运量','实际客运量');9 l; z4 B2 A, h3 c: e
    
39. xlabel('年份');ylabel('货运量、万人');
    5 k8 v; p& A* }2 z
40. title('运用工具箱客运量学习和测试对比图');
    & Q7 D2 W2 z2 U5 y+ Q
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    ! [9 {# Z$ o- Q
42. legend('网络输出货运量','实际货运量');! v8 R7 T* n! ^' U
    
43. xlabel('年份');ylabel('货运量、万吨');5 e; D7 I- d  ]1 k+ m6 j8 q' j( V
    
44. title('运用工具箱货运量学习和测试对比图');! n( O+ L5 O0 C7 V, o
    
45. 1 Y9 P# m3 V$ S2 A5 t4 S
    
46. %新数据仿真/ z4 F1 t1 l/ m( N% T+ [1 G
    
47. pnew=[73.39,75.55) `, D  h8 U. d\" `: _
    
48.      3.9635,4.0975! a7 {. [+ p$ R; z9 C
    
49.      0.9880,1.0268];
    9 z+ p2 B7 A\" j# w: {
50. pnewn=mapminmax('apply',pnew,ps1);& `% A/ F: ^- W5 \. |  p# }& B3 m
    
51. anewn=sim(net,pnewn);
    8 [1 L9 w% b# ?: R! x% 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  T- n/ T9 j: \+ R) i0 |
   
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,...  p) r, Y% H! y) c) F- i4 @
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];$ q/ i+ p& P3 N\" E
   
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 ~! S- l\" ]: U+ @% y; W
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];8 E* E4 x* k8 f8 _
   
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,...# y7 k4 f6 X# H& ]
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   # b$ m( I# o: D( C5 C+ U  g
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...3 G! r% z. i\" I! l0 M
   
9.      22598,25107,33442,36836,40548,42927,43462];4 C4 ]3 X5 A, ]2 R( ]% u
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,..., o/ P2 m8 b( v
    
11.      13320,16762,18673,20724,20803,21804];2 v! J2 x' M# B. z0 U' q: o
    
12. p=[sqrs;sqjdcs;sqglmj];1 o2 _9 n8 r: i& E
    
13. t=[glkyl;glhyl];
    5 s- W, [6 o& A5 q7 ~
14. %数据归一化8 l$ b2 I& L( g( |8 M
    
15. [pn,ps1]=mapminmax(p);
    9 L- Y' z2 Z4 a; s
16. [tn,ps2]=mapminmax(t);7 O3 X; S* I8 ~6 A& L! k  _8 Y
    
17. %dx=[-1,1;-1,1;-1,1];# X, G9 Z+ ~9 U\" }
    
18. p0=minmax(pn);t0=minmax(tn);
    0 r\" b; `' f5 A
19. 
    3 M! N0 g. t9 {* I
20. %BP网络训练# A\" `; N\" J6 j3 G5 a+ x
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); 3 s4 q  d1 K0 y5 w
    
22. net.trainParam.show=1000;: r7 i) y1 b5 R+ @! _, X
    
23. net.trainParam.Lr=0.05;
    ( g: l( X4 P& \* U- P, V- r+ L
24. net.trainParam.epochs=50000;4 E, [) z* l8 {% R$ _6 D2 ^
    
25. net.trainParam.goal=0.65*10^(-5);
    . k8 n) q$ t3 f0 R$ d
26. net=train(net,pn,tn);* [% |! \\" ~8 {, _) I0 E1 f\" f5 A
    
27. 
    ; S( g# T. t3 q% E7 ^7 V
28. %利用原数据对BP网络仿真
    ) b1 E7 P, K# b4 o; L
29. an=sim(net,pn);
    ' B1 f1 F9 D2 d% Z
30. a=mapminmax('reverse',an,ps2);
    ) I( U/ m# G2 n6 K. x% v3 h. W, t
31. 
    ' s+ ~, ^0 s5 s+ M' |' \
32. %仿真结果与原数据对比测试 + V0 V% o) J% M\" W5 }, t6 X6 f2 t
    
33. x=1990:2009;$ o9 F1 W: O4 k! Y0 y( i
    
34. newk=a(1,:);: N5 P# Y2 `! l- s; ]
    
35. newh=a(2,:);
    / ?) ]+ ?7 }# f% j. J
36. figure(2);
    9 M1 o, q( E/ ?% C! T4 `( T
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    ! U$ i1 v! r; ?0 [1 Z) e
38. legend('网络输出客运量','实际客运量');* j/ ^: X& E; g5 K2 w; E
    
39. xlabel('年份');ylabel('货运量、万人');* {/ m1 I4 b- P0 B/ E' O; z8 y
    
40. title('运用工具箱客运量学习和测试对比图');4 a: f' z+ o1 @8 t) j, b2 C) H7 T* A
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    ! D7 s\" e2 x7 J- ^  c
42. legend('网络输出货运量','实际货运量');
    * w) V9 C; C, H( o8 s
43. xlabel('年份');ylabel('货运量、万吨');8 l7 j- _8 d! \( `- j! K9 r7 |6 Y
    
44. title('运用工具箱货运量学习和测试对比图');% x! n9 m, K3 a) P
    
45. 
      {3 n8 w7 f+ y0 w  A9 h6 t4 U% z: v
46. %新数据仿真
    $ X\" k9 Z3 p9 V! E
47. pnew=[73.39,75.55
    ' Q  E, ^, ^! U# X5 A
48.      3.9635,4.0975\" X9 Z4 J; o. b\" Y
    
49.      0.9880,1.0268];
    7 O$ G% K5 R/ Q, ~6 C  N( `% K
50. pnewn=mapminmax('apply',pnew,ps1);) v- {& M& `% @  y) D4 V- L
    
51. anewn=sim(net,pnewn);
    & I' G* f+ \0 f( _5 M\" N7 }
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
   + Q2 k* W- g, `\" R- A! ?3 _% s
2. net=newff(p0,t0,6,{'tansig'},'traingd');
   7 c# i# M+ I5 h9 I) ~  g
3. net.trainParam.show=1000;3 G  u# q. n' G) z6 v
   
4. net.trainParam.Lr=0.05;
   9 F% e, g  W/ `0 v# |( F- V
5. net.trainParam.epochs=2000;
   & r9 [$ n% q# ~# p
6. net.trainParam.goal=0.65*10^(-4);& d* E/ O+ I$ M- G
   
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
" X: F5 g3 F# h  f9 C7 ]) t二、训练函数改成梯度下降BP算法 traingd
6 f$ q  r/ J; N( ]0 p三、迭代次数改成2000
( X" y8 Y# T5 y; }; T0 U1 p上面的参数是自己试的，我也不知道为什么。
( C# _+ a1 y" A8 o" @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。
4 l1 v# o. M# R1 B0 K7 x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