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
4 ~! g0 H& g4 f. n, w+ ~' ]" ?* aclc
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,...
. |2 z. m' z; g0 R$ y  {    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,...
. |5 p- U' k! I! M6 ]    2.5,2.6,2.7,2.85,2.95,3.10];
: j2 z. j1 }% x$ T! Q2 \- dsqglmj=[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];
' J2 s! q9 k' Z) m2 q% Eglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
    22598,25107,33442,36836,40548,42927,43462];
" H5 S* {' _0 {- M+ S4 rglhyl=[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];
- j8 {4 w7 `! @. H3 R- J
8 p! `/ O2 w; W( W: D5 ^9 @%数据归一化
[pn,ps1]=mapminmax(p);
1 x8 C( Z" ^# M5 a[tn,ps2]=mapminmax(t);
dx=[-1,1;-1,1;-1,1];
" I% D' b# [  E) A+ I
%BP网络训练
  X& D0 n5 L1 v6 Inet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
/ ?( O' F1 u' y& p# `* Xnet.trainParam.Lr=0.05;
$ A0 v; l- o; P) ?" ~net.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
& d, {) e' }/ u/ A& S3 _2 o, u
9 @7 C* Y- B* w3 _& D* T% H%利用原数据对BP网络仿真
an=sim(net,pn);
$ p5 F, [/ d4 Z& na=mapminmax('reverse',an,ps2);
" f0 Y' x- o% N# o% \
$ t: v, [. P9 t; x+ O1 n7 E%仿真结果与原数据对比测试
x=1990:2009;
- u7 [* R6 a" Q- e- `+ \0 Tnewk=a(1,:);
newh=a(2,:);
/ y) K% _" L6 F( e. K4 s( _figure(2);
3 n) e- T% _- J( p9 H0 Psubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
6 w* |/ d6 R- \4 N, Y$ y5 W  ^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');
4 L/ o$ }. f/ ^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');
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
7 D% Y, F& U* ^2 A$ M; D    3.9635,4.0975
    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
% S9 b2 _! f+ v9 s: kanew=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 ...

5 G" j. D/ B4 C) V1 _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');
+ f, a# z* ?; X' h, ?我修改为：
4 b2 D! p3 G' \$ Q7 y+ U2 anet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
: g2 D/ }& w: [: m8 @9 g3 w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：
5 m* z( ?' ]4 ], g. R) e[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
/ w% f8 y! |( [! |a=postmnmx(an,mint,maxt);
pnewn=tramnmx(pnew,minp,maxp);
, v" }& z. w, R1 Yanew=postmnmx(anewn,mint,maxt)
9 B3 x* o- a7 ^* B+ i3 i我修改为：
4 j. J$ ^3 Y* w" s[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
a=mapminmax('reverse',an,ps2);
pnewn=mapminmax('apply',pnew,ps1);
anew=mapminmax('reverse',anewn,ps2)

1 v+ m5 y  @# @/ M- G原程序为：
( I. u* b6 Q" w%原始数据输入
0 I7 g6 M! j( w) `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,...
0 j+ h+ m# j/ _- S5 q6 H4 Z2 l    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
  ?' h' }$ `$ l& u: ^+ msqjdcs=[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];
# V: V  Z# ^$ c' Cglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
3 h  Y5 k7 v- m& D' c0 q+ L    22598,25107,33442,36836,40548,42927,43462];
# @( G5 t9 n# T  D8 l3 M; Eglhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
6 T: d( }) ]3 P5 c    13320,16762,18673,20724,20803,21804];
4 X; M: _, n/ A9 x6 a+ Rp=[sqrs;sqjdcs;sqglmj];
1 J! J7 r2 G1 C, T; St=[glkyl;glhyl];

: u$ i5 q2 N8 K1 S& r$ s%数据归一化
5 P9 e2 {$ u1 T; B# x' h[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
dx=[-1,1;-1,1;-1,1];
1 U1 _0 ~; A3 k$ b! n0 E
; C) P$ J0 ]6 C, T; c%BP网络训练
net=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
, Y; u0 D5 u/ U: @2 b( d# z( Znet.trainParam.Lr=0.05;
; U6 s6 D  Y) E0 N- X7 b( I4 ~' @net.trainParam.epochs=50000;
( I9 @' O  b1 x/ N/ j6 a* r! Ynet.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
- ?# x' y+ u& 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
5 a/ T/ e/ E: cx=1990:2009;
6 g% u" r$ p4 F' s/ xnewk=a(1,;
newh=a(2,;
1 D/ W; r' i9 [+ s8 m; Z) lfigure(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');
; Y8 u/ B0 K- ~subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
% Q; V4 l+ q8 K! x7 k1 U" |4 R8 F9 @legend('网络输出货运量','实际货运量');
+ k( r6 I- w6 a8 z, S+ J7 cxlabel('年份');ylabel('货运量、万吨');
, o9 r! j# A4 F5 {2 n  Stitle('运用工具箱货运量学习和测试对比图');
! q! ?1 W, l8 C, ]9 n
%新数据仿真
pnew=[73.39,75.55
4 C- h1 ^9 a( x  Z. h) O2 V    3.9635,4.0975
# c& `3 F5 e4 Q4 K# |: [  v1 [. ]    0.9880,1.0268];
pnewn=tramnmx(pnew,minp,maxp);
anewn=sim(net,pnewn);
8 r9 [& n2 g" y3 K1 @anew=postmnmx(anewn,mint,maxt)

修改后程序为：
%原始数据输入
% h: W* k, S9 a! A4 x* _) Iclc
  |9 F3 V& E( {; y: T" y, }5 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];
' o8 c# E' {6 b( m2 Bsqjdcs=[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];
# x' M4 l% ~8 s. q* j0 Osqglmj=[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];
; K: ~1 T4 {; x1 Q' g. ?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];
: E9 f2 `) i" q- H* g7 P8 gt=[glkyl;glhyl];
. b. K$ x; F" b
1 m) v- y, |" m& }% |. w) @+ p%数据归一化
& E/ }  c9 Y, D& {; {/ e[pn,ps1]=mapminmax(p);
[tn,ps2]=mapminmax(t);

' t0 O4 ~& h, f%BP网络训练
3 a; A7 v, ^9 S5 p# n. Wnet=newff(p,t,[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);
net=train(net,pn,tn);

%利用原数据对BP网络仿真
% v7 b0 L! V$ w2 n% x* _an=sim(net,pn);
. F% Y% `; _9 Ka=mapminmax('reverse',an,ps2);
3 L9 p2 m  D9 D$ g1 U7 T
%仿真结果与原数据对比测试
  ]0 u6 O: S* Nx=1990:2009;
newk=a(1,;
newh=a(2,;
figure(2);
8 w% Y% }) n! hsubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
legend('网络输出客运量','实际客运量');
* o9 e# [0 f: t- o5 Hxlabel('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');
- D, {$ u9 \. A, g: ~xlabel('年份');ylabel('货运量、万吨');
1 a6 e: Z# l& l0 B! [% ~title('运用工具箱货运量学习和测试对比图');
2 }$ P% U; q, Y9 h
+ q' `* `6 s. D0 O%新数据仿真
pnew=[73.39,75.55
    3.9635,4.0975
# u+ B2 T3 O' c- F    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
anewn=sim(net,pnewn);
. D! O! k4 a* eanew=mapminmax('reverse',anewn,ps2)
（修改的地方用颜色标记了）
3 m$ D% C& C  Z9 N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
   % }, X2 F* k: b. T9 _3 d
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: @8 }% S0 Y$ y( J& a6 q
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   % k4 [3 G9 @+ R9 T2 d4 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,...
   ( L( b4 w/ ^3 ^( ?8 q- j3 W7 U
5.     2.5,2.6,2.7,2.85,2.95,3.10];% S( _, h! W9 P+ q1 q; R
   
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,...) h4 N+ {% s9 T
   
7.     0.56,0.59,0.59,0.67,0.69,0.79];# ~' l) B: {5 {1 P; z2 @6 r/ v
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   % ?# F8 x7 z) b
9.      22598,25107,33442,36836,40548,42927,43462];# Z5 U* n  h1 Q1 C, t( s
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...  F4 u. ^/ G' r\" u3 H% Z
    
11.      13320,16762,18673,20724,20803,21804];\" Z& I2 o8 w+ r% p. @, S- L2 C1 i
    
12. p=[sqrs;sqjdcs;sqglmj];
    # {/ R; |; v) c3 f2 B
13. t=[glkyl;glhyl];
    1 p% h7 Y3 j# c  E\" a. S5 ?* u
14. %数据归一化
    ( R5 Q9 d; G2 b5 b
15. [pn,ps1]=mapminmax(p);
    6 P3 Z, o# Z) r2 U
16. [tn,ps2]=mapminmax(t);
    ' Z& t' H3 M8 H/ M4 m+ }  I+ }
17. %dx=[-1,1;-1,1;-1,1];) j; _* G- l  s% Z7 ?+ t/ d9 i
    
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]
    * u( i; y( n( q  I
19. ' w: w$ b1 r% [0 b
    
20. %BP网络训练
    5 i# I% K, \3 A2 `8 N
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]4 A1 y' c' d' V6 A
    
22. net.trainParam.show=1000;
    - ~1 X+ \7 X. d6 c
23. net.trainParam.Lr=0.05;, ?8 D1 d4 f% T# r
    
24. net.trainParam.epochs=50000;
    5 ?. N  p+ `$ o  e4 @
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]
    0 j$ d9 s2 F8 x& t: v
26. net=train(net,pn,tn);\" ]# i% Z+ X0 V5 G3 e3 H: H
    
27. ( T6 {$ k( I) `# M  c& _1 D- g
    
28. %利用原数据对BP网络仿真
    7 M' h. G4 Q. |, M! `  |2 |( n
29. an=sim(net,pn);
    ) r  [7 G  q! `$ m
30. a=mapminmax('reverse',an,ps2);5 R, z/ Z/ T1 U
    
31. 
    6 @5 M0 x1 V* p2 Y* o\" z! o
32. %仿真结果与原数据对比测试
    , }6 A/ z; V: P\" R
33. x=1990:2009;/ i. f( B0 q# h
    
34. newk=a(1,:);
    : N: |/ R- ^5 v$ a/ V6 |  u) _
35. newh=a(2,:);: e4 t; A$ K\" Q  o# |\" T3 ]\" ^* j
    
36. figure(2);
    $ ^8 u9 b* U5 S
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    0 a* J1 }2 F; A. R% i8 g. j! y
38. legend('网络输出客运量','实际客运量');
      ~  L% y/ S. l9 g3 j
39. xlabel('年份');ylabel('货运量、万人');
      ]+ E# Q! d8 K+ |9 y/ \
40. title('运用工具箱客运量学习和测试对比图');9 z, y# M. a6 ~5 s  `7 {2 O
    
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    ' B3 X- E9 h: k7 X
42. legend('网络输出货运量','实际货运量');6 E- r; m8 ]4 P
    
43. xlabel('年份');ylabel('货运量、万吨');
    ) |' @2 N3 S1 F' D0 w3 @
44. title('运用工具箱货运量学习和测试对比图');
    8 I- J0 u+ ?7 n: Q  g
45. 1 Y' L# p! F( W8 z
    
46. %新数据仿真
    ; z) B/ Z2 x; v/ R4 m( ?
47. pnew=[73.39,75.55
    % p) a2 Q7 K. p+ z
48.      3.9635,4.0975
    8 O$ ?% r  Y% q- d4 ]4 K3 e: B
49.      0.9880,1.0268];
    & |0 l# _, T\" k$ Y! d  [5 g
50. pnewn=mapminmax('apply',pnew,ps1);
    9 }0 B5 p7 |4 I! Z) s
51. anewn=sim(net,pnewn);
    8 \' s1 r) N2 J7 p- x/ Q* F
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. clc0 l2 L# ?. T8 Y) v3 y4 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,...0 e: V2 s. I( s: T0 O* ^
   
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];$ _- J+ `) U9 t
   
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,...$ l) X1 F\" s/ ^\" O) E1 ~6 Q
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];( m9 S8 Z; w\" \
   
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,...
   ) }\" I6 `# o# O. f) |6 r5 f
7.     0.56,0.59,0.59,0.67,0.69,0.79];, O6 L# H6 m! ^+ @7 B\" W( a3 r9 Q& o
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...$ e2 Y\" x1 t1 `\" y& W; q* ]2 t9 b
   
9.      22598,25107,33442,36836,40548,42927,43462];9 }- ~0 ]! Y# Q- ~+ Q# \
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,.... q1 q9 [; `( k* k
    
11.      13320,16762,18673,20724,20803,21804];% D7 q* h' q+ V  Z1 |
    
12. p=[sqrs;sqjdcs;sqglmj];
    % c7 Y7 Z8 B2 E  N
13. t=[glkyl;glhyl];
    $ u8 z6 a5 c9 Z! ]
14. %数据归一化
    : ]% D: i2 n5 d$ K+ J  e
15. [pn,ps1]=mapminmax(p);
    7 i/ V$ _# n) S3 H1 F4 _
16. [tn,ps2]=mapminmax(t);
    # s' g0 `  `\" y: n1 q
17. %dx=[-1,1;-1,1;-1,1];
    8 N\" X$ y* ^* l
18. p0=minmax(pn);t0=minmax(tn);
      K/ K0 {2 N4 y  R; x. P
19. 
    * o8 d0 |+ M2 c4 K6 }7 g7 p1 M
20. %BP网络训练9 M' R. a4 |\" b( L, G6 N, ^. E# n
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
    1 Y' v4 N9 {  j9 `4 D
22. net.trainParam.show=1000;9 T0 e% z+ i' d
    
23. net.trainParam.Lr=0.05;. W% Z& m2 h0 [) d4 o
    
24. net.trainParam.epochs=50000;
    * }' ?\" [- M: @- N\" m
25. net.trainParam.goal=0.65*10^(-5);
    & s/ n) `6 J\" {  l- s- G
26. net=train(net,pn,tn);
    % T& P) k8 J: @9 {
27. 
    ) ?; j( [2 K! d; v) ^
28. %利用原数据对BP网络仿真
    ) C) F3 c: ^8 h( F
29. an=sim(net,pn);1 @9 l7 b: j$ t% t& ?2 B* |
    
30. a=mapminmax('reverse',an,ps2);
    ) D7 G6 J; ?9 d3 J8 {\" _8 R& {
31. 
    ; e$ Y! ?7 Y+ X  B7 d
32. %仿真结果与原数据对比测试 2 Y5 l, u' n& t7 o( J; v
    
33. x=1990:2009;
    7 E+ R4 A( Y  ?+ U9 l5 i# g
34. newk=a(1,:);# \' `% f' {+ r, z7 `! Q- j
    
35. newh=a(2,:);) r2 t+ ?! ]% Q: `: o; C
    
36. figure(2);4 l) q\" T  K: b) U8 A5 u
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');! n6 `& y; S2 _! l% [  x& E
    
38. legend('网络输出客运量','实际客运量');\" B8 M3 A4 g% _
    
39. xlabel('年份');ylabel('货运量、万人');4 U\" X\" M! z4 O  G$ c( _3 O\" h4 n
    
40. title('运用工具箱客运量学习和测试对比图');
    6 ^% i7 o- W  X: f, h. c\" }  E
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    . m7 Y$ @  p& t% b, R3 n6 X
42. legend('网络输出货运量','实际货运量');
    \" w# j\" B  Z8 n3 W# `( e
43. xlabel('年份');ylabel('货运量、万吨');
    4 R\" i* {# |5 C8 e4 a. T& H
44. title('运用工具箱货运量学习和测试对比图');
    % ~: C2 |) S\" d0 W) ~7 N0 ?
45. ' t% ?% r( S- X. P+ b; x- c
    
46. %新数据仿真' T8 F$ _5 `+ G% `, D1 a/ A+ t0 M: n
    
47. pnew=[73.39,75.55
    9 H4 a$ ~: B! M0 I$ d
48.      3.9635,4.0975
    - A1 N% }$ q  m$ S
49.      0.9880,1.0268];9 T2 K1 f2 \. @
    
50. pnewn=mapminmax('apply',pnew,ps1);( {2 N; s3 y6 B/ P9 u' ^/ q( w
    
51. anewn=sim(net,pnewn);
    3 S; u' I5 v\" q+ Z/ D2 T! e' L$ ^
52. anew=mapminmax('reverse',anewn,ps2)

复制代码

T-Eric

且生 发表于 2014-7-21 18:22
1 x  Q. e; c4 R# ]* G# \' T0 b# |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
   - e8 f7 O; s' f, Q* \! }; c7 j4 r
2. net=newff(p0,t0,6,{'tansig'},'traingd');
   : z! y6 _$ I5 `; }+ f
3. net.trainParam.show=1000;
   4 ~4 }. `& U7 K' v& ~) a$ N4 U2 M
4. net.trainParam.Lr=0.05;
   5 X6 N7 ~* {' J$ t
5. net.trainParam.epochs=2000;
   + [) M+ t/ {0 p+ k! U) e3 O' ?
6. net.trainParam.goal=0.65*10^(-4);
   - b' ]' N; K+ {# q
7. net=train(net,pn,tn);

复制代码

我只改了这里面的，
4 Y) @8 b. |% h! A) d3 z7 {, ?一、改成单隐含层的，6个节点
1 l) @; c3 A+ H6 X& `2 q/ M二、训练函数改成梯度下降BP算法 traingd
三、迭代次数改成2000
3 o) U$ F8 K* z* K) f5 ?上面的参数是自己试的，我也不知道为什么。
3 E4 r, c  C/ y+ _7 [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