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！！
. g8 h  ^! A3 }' W
4 {4 Z  S$ P) Q# K%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,...
( }" y, b+ f3 E$ z2 E- [    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,...
. z# j0 G+ U. R; 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];
- K% i* {% p  F' B& @3 O% Dglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
5 L! d$ A' |9 e7 A8 X$ v- o    22598,25107,33442,36836,40548,42927,43462];
glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
: Y% n+ x' Z& b0 ?6 R    13320,16762,18673,20724,20803,21804];
! B2 m6 C( a9 v% q0 W( e! p/ Op=[sqrs;sqjdcs;sqglmj];
t=[glkyl;glhyl];
! {) B  G  h  K
+ t# X% {6 S. t) a8 |) N" |%数据归一化
[pn,ps1]=mapminmax(p);
' L! S2 P$ m# I: {[tn,ps2]=mapminmax(t);
dx=[-1,1;-1,1;-1,1];
) i3 J! {. }- O! O  W- G+ q
( Y4 Q$ t% D3 R%BP网络训练
- R3 L0 o' {5 W* s/ L; Z# }, |+ Ynet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
. b; D& P% ~) A; wnet.trainParam.show=1000;
) u8 Y+ V) |( \! V! t* cnet.trainParam.Lr=0.05;
; v% c6 M; ]( z( s) Q& U7 V; l5 ~net.trainParam.epochs=50000;
& Q. K0 s7 {$ l0 U% g* x9 anet.trainParam.goal=0.65*10^(-3);
net=train(net,pn,tn);
, ]2 r. y5 x, q( _' q) a* D
, G' n  U% i( z%利用原数据对BP网络仿真
an=sim(net,pn);
1 z- I. V- b$ H; Y: ^1 b0 Q% ca=mapminmax('reverse',an,ps2);

; l( S# S& R$ t3 c" b' |8 U' H%仿真结果与原数据对比测试
# \2 E. V. b8 o5 ax=1990:2009;
( o- ?  u) G. Y8 Q, B2 `5 gnewk=a(1,:);
5 Z0 p# \3 d3 E6 |6 i! inewh=a(2,:);
5 R( S/ \4 E, R" L9 u# v' J& b8 Qfigure(2);
% x9 j2 [  ?: E$ ssubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
legend('网络输出客运量','实际客运量');
6 W$ `& I, y$ M) o0 Xxlabel('年份');ylabel('货运量、万人');
4 _' t+ M. F6 h# l0 X' S6 v* D4 V& T. Ytitle('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--+');
3 c- F. L2 N. k& R1 F* x! I3 Nlegend('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');
. k  J5 E6 t. U0 y
, T$ g$ t. L0 l1 p, }5 S) c9 {%新数据仿真
% [% d3 \: J' Wpnew=[73.39,75.55
' b4 c. W0 Y% q' o    3.9635,4.0975
    0.9880,1.0268];
" L- c0 b9 _" P9 |2 ^4 o& @pnewn=mapminmax('apply',pnew,ps1);
2 S7 [" C0 d" P. }- ?anewn=sim(net,pnewn);
anew=mapminmax('reverse',anewn,ps2)
0 f/ E1 O/ h( S8 v% E. x/ N! U

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
* n. w* }# H; u4 b, |+ E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：
net=newff(dx,[3,7,2]{'tansig','tansig','purelin'},'traingdx');
( m) |2 F3 ?- g我修改为：
net=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
8 b5 d5 L) C) E3 A- c9 a0 R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：
% ]% D, l: k% D. _; z! M# {( Q[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
' B. P" m" b+ w: @# Z& p% Ta=postmnmx(an,mint,maxt);
pnewn=tramnmx(pnew,minp,maxp);
$ \9 X9 a1 t. y# d6 O( Uanew=postmnmx(anewn,mint,maxt)
* S9 E, `4 C. \: E我修改为：
0 S# ^& f; h' D( Q* P: A[pn,ps1]=mapminmax(p);[tn,ps2]=mapminmax(t);
a=mapminmax('reverse',an,ps2);
pnewn=mapminmax('apply',pnew,ps1);
- h7 F' i' b( w- ~  ^& @anew=mapminmax('reverse',anewn,ps2)
1 l: u# b+ e5 V/ E
' `% V. `& F' [5 q( A原程序为：
%原始数据输入
; k& i2 x7 V  a) Wclc
+ h' K: F: i' Z5 a0 T6 b3 Ysqrs=[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];
% t% c0 y4 \: J) D4 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,...
0 K& i5 Y7 g. s- p6 x8 A8 C6 M    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,...
$ E) v- Q$ D) k& Q6 E3 A    0.56,0.59,0.59,0.67,0.69,0.79];
* j1 x) I) U. t: ~7 Q3 qglkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
! ~3 T$ W3 m' w. H    22598,25107,33442,36836,40548,42927,43462];
glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
' P5 G  z; `- O* q: w5 s" n% @% a    13320,16762,18673,20724,20803,21804];
p=[sqrs;sqjdcs;sqglmj];
# j4 X  w% D/ I, X' Ot=[glkyl;glhyl];
: @' j4 Z0 o# n$ H! R9 S
%数据归一化
[pn,minp,maxp,tn,mint,maxt]=premnmx(p,t);
% Y7 y- n& M0 |4 Rdx=[-1,1;-1,1;-1,1];
4 g0 D4 B$ i% [: z* l
$ p! n! I' ^- S( v8 x%BP网络训练
net=newff(dx,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
3 c' `* T1 ^2 |3 F$ I2 \net.trainParam.Lr=0.05;
& i& v- z) Q, M8 pnet.trainParam.epochs=50000;
net.trainParam.goal=0.65*10^(-3);
) @* N3 X8 I! W0 unet=train(net,pn,tn);

%利用原数据对BP网络仿真
0 N* `0 M$ D. [: |: |1 y0 ]an=sim(net,pn);
a=postmnmx(an,mint,maxt);

%仿真结果与原数据对比测试
( B3 E0 m0 B" o* H$ Y2 Hx=1990:2009;
0 I7 U& [0 G7 X& f0 ]% Unewk=a(1,;
9 a: w/ ]2 F9 ^! N0 w% Qnewh=a(2,;
figure(2);
subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
. \, n" e/ ~: A+ j+ ?7 Ylegend('网络输出客运量','实际客运量');
( l) f3 e% o5 T( c. _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');
- F7 b( p( k) Q% y! hsubplot(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
# D( x, s' M3 g9 ~1 Qpnew=[73.39,75.55
3 r6 F. t, @) Z8 L    3.9635,4.0975
& |9 w8 u: j9 f. @& R    0.9880,1.0268];
* v# g& S! _6 h% q# h" f4 Hpnewn=tramnmx(pnew,minp,maxp);
anewn=sim(net,pnewn);
8 J& _" C$ @7 ?" D  f$ M1 janew=postmnmx(anewn,mint,maxt)
# i/ X3 `/ P+ |& }
5 h6 L% G2 M1 ~6 d4 [% \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,...
    41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
/ L; p. ?# m% x1 y; Q; n7 Rsqjdcs=[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,...
3 ?  e" h1 j0 S    2.5,2.6,2.7,2.85,2.95,3.10];
6 a. g& [, m* F6 ]7 ]3 G" V1 b6 c$ ^4 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];
9 B4 k. _- t0 j" r& H" Xglkyl=[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];
6 }+ ?( }- S. X% T% S. F
%数据归一化
4 n  p4 a# U' x* c9 @; ~[pn,ps1]=mapminmax(p);
7 P1 S' a* B& A9 T, Y[tn,ps2]=mapminmax(t);

% a3 u) Q2 u+ Z8 m1 l4 C" ^%BP网络训练
- U/ h  ]! a7 U& Y# Xnet=newff(p,t,[3,7,2],{'tansig','tansig','purelin'},'traingdx');
net.trainParam.show=1000;
3 `, s! T" |8 ]( V. pnet.trainParam.Lr=0.05;
& Y: v5 P& L( i( {net.trainParam.epochs=50000;
8 G2 ^1 A9 t3 ^2 T9 G) o( j% Dnet.trainParam.goal=0.65*10^(-3);
/ |( q" O+ L4 ?6 p1 s( e# ]0 |net=train(net,pn,tn);
6 W5 V2 w* P( r' ?6 x
6 W$ m; b* e/ T, ^( E%利用原数据对BP网络仿真
' S0 g# t* H+ _: @0 x! aan=sim(net,pn);
0 W% j- z% M% Y& D. _' x0 ^, }& wa=mapminmax('reverse',an,ps2);
" ?: f1 f6 T6 r- f
: U: z! t8 V3 N5 j" P2 _" G%仿真结果与原数据对比测试
x=1990:2009;
4 o1 Q1 `/ `9 R) jnewk=a(1,;
newh=a(2,;
+ p/ _" E0 X/ |0 n2 _: bfigure(2);
2 Q' v& H2 F7 A" Z$ g/ d3 U5 P" ssubplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
. K- }/ c0 b% H8 e6 alegend('网络输出客运量','实际客运量');
7 N" D1 F, |7 J$ J$ Uxlabel('年份');ylabel('货运量、万人');
2 _& H7 Y# {9 T4 }$ Ltitle('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--+');
; `9 C' _+ o: u4 Q9 N4 klegend('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');
# b3 e9 G+ a7 P1 I" `
%新数据仿真
pnew=[73.39,75.55
* ^& M6 b/ S+ e    3.9635,4.0975
& \7 R5 w! r/ [1 D6 J" G    0.9880,1.0268];
pnewn=mapminmax('apply',pnew,ps1);
2 g6 y' E7 A* q: m: vanewn=sim(net,pnewn);
1 p, V3 `) [. _anew=mapminmax('reverse',anewn,ps2)
8 ]5 I/ W& [2 m- s（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
   * [1 q8 L  p. m: Z8 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,...
   0 ^$ e* P; Q/ O) S) ^( b0 d
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];
   \" h0 t8 @9 v6 m, n
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,...  A* ^; `  ~9 d\" x\" X, |
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   \" T% A+ H/ {1 y2 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,...
   ( [6 ~# X- g5 H& s\" c
7.     0.56,0.59,0.59,0.67,0.69,0.79];4 d; w5 e; P3 X) s/ ^$ p
   
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...
   : n' A( o. U4 i, }3 G
9.      22598,25107,33442,36836,40548,42927,43462];) t/ }\" E5 S- y
   
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...0 o3 E. s7 e5 v8 O
    
11.      13320,16762,18673,20724,20803,21804];
    - Q5 f' ~- Z  q, f' h
12. p=[sqrs;sqjdcs;sqglmj];
    ( c% T% N4 Q+ M/ ~- a
13. t=[glkyl;glhyl];
    + c- [# X* v$ k6 _* N6 q
14. %数据归一化
    / e\" `) B+ h8 w\" w' {
15. [pn,ps1]=mapminmax(p);* [% t% X# _, J9 D: b
    
16. [tn,ps2]=mapminmax(t);9 c9 m+ J+ z* u# C
    
17. %dx=[-1,1;-1,1;-1,1];& o6 d* d5 T1 I4 ?( u+ V6 n) n
    
18. [color=Red] p0=minmax(pn);t0=minmax(tn);[/color]
    , }) y  \- ]3 |* E: r% `8 e
19. 7 b$ v* x4 O: W
    
20. %BP网络训练
    + {9 v6 O9 B0 \7 w1 F4 @
21. [color=Red] net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); [/color]0 S( J9 {7 Z9 g
    
22. net.trainParam.show=1000;% u- a* o) ]7 U& C; M! _  |+ @
    
23. net.trainParam.Lr=0.05;
    ; L\" b% J% |, G$ q8 A
24. net.trainParam.epochs=50000;3 N( h& H5 }; ~2 N
    
25. [color=Red] net.trainParam.goal=0.65*10^(-5);[/color]; @: S8 J7 x3 z
    
26. net=train(net,pn,tn);2 c, K, W* T6 T5 T/ n
    
27. ! A) P5 E  J+ u+ w
    
28. %利用原数据对BP网络仿真. P9 T, x% q/ T# Y6 N
    
29. an=sim(net,pn);% S% G; y! J( q5 d4 ^& U% w
    
30. a=mapminmax('reverse',an,ps2);
    : S) r& s' x: d4 J0 j
31. 8 Y0 Q+ }! H0 D! G4 I- I7 T
    
32. %仿真结果与原数据对比测试
    + `8 s* D1 P* q3 Q7 y$ _5 c
33. x=1990:2009;
    2 [% W- V9 B) J
34. newk=a(1,:);3 h7 p/ n' D. \8 \$ f* G. n1 c
    
35. newh=a(2,:);, ~$ _* l$ P. A8 o8 p, H9 q
    
36. figure(2);
    ) A: Y, P0 |  Q, i! t- \& R
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');4 H* W) N\" @% c1 l5 ]* N& E1 j
    
38. legend('网络输出客运量','实际客运量');
    3 @: t. I/ N2 s/ C
39. xlabel('年份');ylabel('货运量、万人');3 }) h6 B8 {: s; i5 {* k
    
40. title('运用工具箱客运量学习和测试对比图');
    6 e! P7 _! ]8 g8 r
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');
    ' F6 @: d6 r4 `. j\" x+ h0 j% o
42. legend('网络输出货运量','实际货运量');
    : R$ ^7 T# R3 ^& C2 D9 j
43. xlabel('年份');ylabel('货运量、万吨');
    ' ~4 H9 ~. b4 |$ ~( G8 O; f
44. title('运用工具箱货运量学习和测试对比图');( O4 s9 U\" [$ K8 M: U
    
45. 
    & b6 i; c7 K1 b) b. h3 b
46. %新数据仿真8 y  n; P' Y+ l+ Z
    
47. pnew=[73.39,75.55* `: k9 t! V6 L
    
48.      3.9635,4.0975
    4 W8 P6 q% r, p- n. S) M8 D
49.      0.9880,1.0268];) \$ J$ j9 b5 W: c2 s/ K
    
50. pnewn=mapminmax('apply',pnew,ps1);
    ' V2 ]. X$ G\" t7 E% f; _, n9 ]
51. anewn=sim(net,pnewn);' j( y5 c2 J* b; B! \
    
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; k% ]  M3 J6 f0 d& R8 U
   
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,...
   : ?7 J# Q/ Q9 [2 A; P. G( V) l
3.     41.93,44.59,47.30,52.89,55.73,56.76,59.17,60.63];1 b/ p+ i: v8 m8 H/ B5 i
   
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,...; d  {2 p! d$ ~) E
   
5.     2.5,2.6,2.7,2.85,2.95,3.10];
   3 D2 G# d$ 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,...
   3 f9 h5 Y: J% V4 x
7.     0.56,0.59,0.59,0.67,0.69,0.79];
   , E- o+ L5 }8 ^/ B/ F' y
8. glkyl=[5126,6217,7730,9145,10460,11387,12353,15750,18304,19836,21024,19490,20433,...& f% k8 t5 [) K\" r) u! S4 H' F
   
9.      22598,25107,33442,36836,40548,42927,43462];
   ; c: O! n+ u' h
10. glhyl=[1237,1379,1385,1399,1663,1714,1834,4322,8132,8936,11099,11203,10524,11115,...
    . M, ~& C' d\" R5 y+ l% ^+ z* R
11.      13320,16762,18673,20724,20803,21804];
      ^+ j: ?( K9 v) f( w, _
12. p=[sqrs;sqjdcs;sqglmj];
    2 o, T( ^: J; s\" @# E
13. t=[glkyl;glhyl];
    : C1 f2 X3 |$ \+ z
14. %数据归一化  t) p0 y1 I) V\" {6 O. ?
    
15. [pn,ps1]=mapminmax(p);\" W2 e6 R  b# r5 o2 ~! d
    
16. [tn,ps2]=mapminmax(t);! I  l0 P% b\" F
    
17. %dx=[-1,1;-1,1;-1,1];/ ^\" E3 n& J, H4 T: e
    
18. p0=minmax(pn);t0=minmax(tn);9 z. b) z0 }1 ~\" C6 z3 }3 d
    
19. 
    $ w3 f# a( ?\" I. v& x
20. %BP网络训练# c6 J+ H$ v3 S3 O, r/ H
    
21. net=newff(p0,t0,[3,7,2],{'tansig','tansig','purelin'},'traingdx'); : u: L! Q- [- ?/ V* W- y; a
    
22. net.trainParam.show=1000;
    & [% s7 d* S, I, Z& f* P
23. net.trainParam.Lr=0.05;
    1 |3 t9 d( `$ c
24. net.trainParam.epochs=50000;1 T$ t7 ?+ x* A% p\" T8 ?
    
25. net.trainParam.goal=0.65*10^(-5);
    9 ^4 n3 h$ W\" t* Z
26. net=train(net,pn,tn);; U2 ~, U- j: C3 C0 O
    
27. 0 Q( b6 j& O1 G$ K
    
28. %利用原数据对BP网络仿真( K% U- ]3 @# M% R0 p
    
29. an=sim(net,pn);\" m, r( L\" B3 X( d# G
    
30. a=mapminmax('reverse',an,ps2);
    ' t  d3 C\" s7 w6 U/ u% L
31. 
    * S' V$ \# N+ k7 P/ ]$ \7 r; |0 M7 V/ E
32. %仿真结果与原数据对比测试
    ( @4 z8 b: y4 h: j% @* X1 W
33. x=1990:2009;
    ( E5 |\" N9 Q1 |) K4 `+ \
34. newk=a(1,:);+ `\" C1 |% U8 d; F
    
35. newh=a(2,:);
    * ?$ U. i: ~. s+ n& D; e
36. figure(2);% z% \% N# L+ |
    
37. subplot(2,1,1);plot(x,newk,'r-o',x,glkyl,'b--+');
    0 g! n7 J, [% j3 t* Q' j  S
38. legend('网络输出客运量','实际客运量');7 x% d6 I. W) s* p1 W
    
39. xlabel('年份');ylabel('货运量、万人');
    $ I- K' U7 q* a- C. {: p, w. N3 |
40. title('运用工具箱客运量学习和测试对比图');
    1 c7 s1 D/ s; P
41. subplot(2,1,2);plot(x,newh,'r-o',x,glhyl,'b--+');4 ]5 S+ h% t+ o. U2 Z
    
42. legend('网络输出货运量','实际货运量');
    8 L' k4 x8 s6 t8 T& X4 R% b
43. xlabel('年份');ylabel('货运量、万吨');
    / g# P* K8 m/ h5 p0 G) Z/ c4 u& t
44. title('运用工具箱货运量学习和测试对比图');2 T2 J\" y9 I  ^( V4 B* u* D2 K
    
45. \" ~7 V, g3 [; u6 c* t$ ^: Y9 z8 c  C* Y- k
    
46. %新数据仿真
    ! E\" I+ w' N0 Y0 B/ z# n
47. pnew=[73.39,75.55
    % r! k' E& [1 c, \
48.      3.9635,4.0975
    5 r; P+ V. ~  x9 a
49.      0.9880,1.0268];
    % n2 [' U% h; Q( Z' U8 C
50. pnewn=mapminmax('apply',pnew,ps1);1 M- l$ d7 U3 _. I# B4 R( q! p, o
    
51. anewn=sim(net,pnewn);
    , o' d8 U4 Q2 O, _/ i. t
52. anew=mapminmax('reverse',anewn,ps2)

复制代码

T-Eric

且生 发表于 2014-7-21 18:22
6 m4 B. D7 G+ ~6 O3 ~# g: K9 q忽略我上面那个，改动的地方在18，21，25行……求批

8 Z) @' _7 |0 N, h) P- L- M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1 ]8 R- r- l! b2 w# v
   
2. net=newff(p0,t0,6,{'tansig'},'traingd');
   5 ^- s\" y+ n& Z6 |' N  m
3. net.trainParam.show=1000;
   9 O6 @) p( i0 V1 K, z. \$ o
4. net.trainParam.Lr=0.05;0 e  k5 ^\" v\" ?8 n+ a7 J
   
5. net.trainParam.epochs=2000;+ o% V) s8 m2 k1 ^  G6 l$ D) `
   
6. net.trainParam.goal=0.65*10^(-4);# V! g/ f/ b7 F# e8 q4 c# ^
   
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
) [; D# A/ T5 C+ [: h二、训练函数改成梯度下降BP算法 traingd
三、迭代次数改成2000
9 v* {3 c' J" [/ r' u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。
7 ^  ~; n- s, Z% c7 F8 b& ^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