神经网络优化学习程序.docx
- 文档编号:29656787
- 上传时间:2023-07-25
- 格式:DOCX
- 页数:13
- 大小:15.97KB
神经网络优化学习程序.docx
《神经网络优化学习程序.docx》由会员分享,可在线阅读,更多相关《神经网络优化学习程序.docx(13页珍藏版)》请在冰豆网上搜索。
神经网络优化学习程序
神经网络优化学习程序
clc
clear
%参数初始化
xite=0.002;
alfa=0.04;
%网络节点
I=6;%输入节点数
M=10;%隐含节点数
O=1;%输出节点数
%系数初始化
p0=0.3*ones(M,1);p0_1=p0;p0_2=p0_1;
p1=0.3*ones(M,1);p1_1=p1;p1_2=p1_1;
p2=0.3*ones(M,1);p2_1=p2;p2_2=p2_1;
p3=0.3*ones(M,1);p3_1=p3;p3_2=p3_1;
p4=0.3*ones(M,1);p4_1=p4;p4_2=p4_1;
p5=0.3*ones(M,1);p5_1=p5;p5_2=p5_1;
p6=0.3*ones(M,1);p6_1=p6;p6_2=p6_1;
%参数初始化
c=1+rands(M,I);c_1=c;c_2=c_1;
b=1+rands(M,I);b_1=b;b_2=b_1;
maxgen=120;%进化次数
%网络测试数据,并对数据归一化
loaddata1input_trainoutput_traininput_testoutput_test
%选连样本输入输出数据归一化
[inputn,inputps]=mapminmax(input_train);
[outputn,outputps]=mapminmax(output_train);
[n,m]=size(input_train);
%%%%%%%网络训练%%%%%%%%%%%
%循环开始,进化网络
foriii=1:
maxgen
iii
fork=1:
m
x=inputn(:
k);
%输出层结算
fori=1:
I
forj=1:
M
u(i,j)=exp(-(x(i)-c(j,i))^2/b(j,i));
end
end
%模糊规则计算
fori=1:
M
w(i)=u(1,i)*u(2,i)*u(3,i)*u(4,i)*u(5,i)*u(6,i);
end
addw=sum(w);
fori=1:
M
yi(i)=p0_1(i)+p1_1(i)*x
(1)+p2_1(i)*x
(2)+p3_1(i)*x(3)+p4_1(i)*x(4)+p5_1(i)*x(5)+p6_1(i)*x(6);
end
addyw=yi*w';
%网络预测计算
yn(k)=addyw/addw;
e(k)=outputn(k)-yn(k);
%计算p的变化值
d_p=zeros(M,1);
d_p=xite*e(k)*w./addw;
d_p=d_p';
%计算b变化值
d_b=0*b_1;
fori=1:
M
forj=1:
I
d_b(i,j)=xite*e(k)*(yi(i)*addw-addyw)*(x(j)-c(i,j))^2*w(i)/(b(i,j)^2*addw^2);
end
end
%更新c变化值
fori=1:
M
forj=1:
I
d_c(i,j)=xite*e(k)*(yi(i)*addw-addyw)*2*(x(j)-c(i,j))*w(i)/(b(i,j)*addw^2);
end
end
p0=p0_1+d_p+alfa*(p0_1-p0_2);
p1=p1_1+d_p*x
(1)+alfa*(p1_1-p1_2);
p2=p2_1+d_p*x
(2)+alfa*(p2_1-p2_2);
p3=p3_1+d_p*x(3)+alfa*(p3_1-p3_2);
p4=p4_1+d_p*x(4)+alfa*(p4_1-p4_2);
p5=p5_1+d_p*x(5)+alfa*(p5_1-p5_2);
p6=p6_1+d_p*x(6)+alfa*(p6_1-p6_2);
b=b_1+d_b+alfa*(b_1-b_2);
c=c_1+d_c+alfa*(c_1-c_2);
p0_2=p0_1;p0_1=p0;
p1_2=p1_1;p1_1=p1;
p2_2=p2_1;p2_1=p2;
p3_2=p3_1;p3_1=p3;
p4_2=p4_1;p4_1=p4;
p5_2=p5_1;p5_1=p5;
p6_2=p6_1;p6_1=p6;
c_2=c_1;c_1=c;
b_2=b_1;b_1=b;
end
E(iii)=sum(abs(e));
end
figure
(1);
plot(outputn,'r')
holdon
plot(yn,'b')
holdon
plot(outputn-yn,'g');
legend('实际输出','预测输出','误差','fontsize',12)
title('训练数据预测','fontsize',12)
xlabel('样本序号','fontsize',12)
ylabel('水质等级','fontsize',12)
%%%%%%%%%%%网络预测%%%%%%%%%%%%%%%%%%
%数据归一化处理
inputn_test=mapminmax('apply',input_test,inputps);
[n,m]=size(inputn_test)
fork=1:
m
x=inputn_test(:
k);
%计算输出中间层
fori=1:
I
forj=1:
M
u(i,j)=exp(-(x(i)-c(j,i))^2/b(j,i));
end
end
fori=1:
M
w(i)=u(1,i)*u(2,i)*u(3,i)*u(4,i)*u(5,i)*u(6,i);
end
addw=0;
fori=1:
M
addw=addw+w(i);
end
fori=1:
M
yi(i)=p0_1(i)+p1_1(i)*x
(1)+p2_1(i)*x
(2)+p3_1(i)*x(3)+p4_1(i)*x(4)+p5_1(i)*x(5)+p6_1(i)*x(6);
end
addyw=0;
fori=1:
M
addyw=addyw+yi(i)*w(i);
end
%计算输出
yc(k)=addyw/addw;
end
%%%%%%%%%%预测结果反归一化%%%%%%%%%%%%%%%%%%%%%%
test_simu=mapminmax('reverse',yc,outputps);
%作图
figure
(2)
plot(output_test,'r')
holdon
plot(test_simu,'b')
holdon
plot(test_simu-output_test,'g')
legend('实际输出','预测输出','误差','fontsize',12)
title('测试数据预测','fontsize',12)
xlabel('样本序号','fontsize',12)
ylabel('水质等级','fontsize',12)
%%%%%%%%%%%%%%%%江水实际水质预测%%%%%%%%%%%%%%%%%%%%%%%
loaddata2CBA
%%%%%%%%%%%%%%%%%%%%%%%%C厂%%%%%%%%%%%%%%%%%%%%%%%%%
zssz=C;
%数据归一化
inputn_test=mapminmax('apply',zssz,inputps);
[n,m]=size(zssz);
fork=1:
1:
m
x=inputn_test(:
k);
%计算输出中间层
fori=1:
I
forj=1:
M
u(i,j)=exp(-(x(i)-c(j,i))^2/b(j,i));
end
end
fori=1:
M
w(i)=u(1,i)*u(2,i)*u(3,i)*u(4,i)*u(5,i)*u(6,i);
end
addw=0;
fori=1:
M
addw=addw+w(i);
end
fori=1:
M
yi(i)=p0_1(i)+p1_1(i)*x
(1)+p2_1(i)*x
(2)+p3_1(i)*x(3)+p4_1(i)*x(4)+p5_1(i)*x(5)+p6_1(i)*x(6);
end
addyw=0;
fori=1:
M
addyw=addyw+yi(i)*w(i);
end
%计算输出
szzb(k)=addyw/addw;
end
szzbz1=mapminmax('reverse',szzb,outputps);
fori=1:
m
ifszzbz1(i)<=1.5
szpj1(i)=1;
elseifszzbz1(i)>1.5&&szzbz1(i)<=2.5
szpj1(i)=2;
elseifszzbz1(i)>2.5&&szzbz1(i)<=3.5
szpj1(i)=3;
elseifszzbz1(i)>3.5&&szzbz1(i)<=4.5
szpj1(i)=4;
else
szpj1(i)=5;
end
end
%%%%%%%%%%%%%%%%%%%%%%B厂%%%%%%%%%%%%%%%%%%%%%%%%
zssz=B;
inputn_test=mapminmax('apply',zssz,inputps);
[n,m]=size(zssz);
fork=1:
1:
m
x=inputn_test(:
k);
%计算输出中间层
fori=1:
I
forj=1:
M
u(i,j)=exp(-(x(i)-c(j,i))^2/b(j,i));
end
end
fori=1:
M
w(i)=u(1,i)*u(2,i)*u(3,i)*u(4,i)*u(5,i)*u(6,i);
end
addw=0;
fori=1:
M
addw=addw+w(i);
end
fori=1:
M
yi(i)=p0_1(i)+p1_1(i)*x
(1)+p2_1(i)*x
(2)+p3_1(i)*x(3)+p4_1(i)*x(4)+p5_1(i)*x(5)+p6_1(i)*x(6);
end
addyw=0;
fori=1:
M
addyw=addyw+yi(i)*w(i);
end
%计算输出
szzb(k)=addyw/addw;
end
szzbz2=mapminmax('reverse',szzb,outputps);
fori=1:
m
ifszzbz2(i)<=1.5
szpj2(i)=1;
elseifszzbz2(i)>1.5&&szzbz2(i)<=2.5
szpj2(i)=2;
elseifszzbz2(i)>2.5&&szzbz2(i)<=3.5
szpj2(i)=3;
elseifszzbz2(i)>3.5&&szzbz2(i)<=4.5
szpj2(i)=4;
else
szpj2(i)=5;
end
end
%%%%%%%%%%%%%%%%%%%%%%%大溪沟水厂%%%%%%%%%%%%%%%%%%%%%%%
zssz=A;
inputn_test=mapminmax('apply',zssz,inputps);
[n,m]=size(zssz);
fork=1:
1:
m
x=inputn_test(:
k);
%计算输出中间层
fori=1:
I
forj=1:
M
u(i,j)=exp(-(x(i)-c(j,i))^2/b(j,i));
end
end
fori=1:
M
w(i)=u(1,i)*u(2,i)*u(3,i)*u(4,i)*u(5,i)*u(6,i);
end
addw=0;
fori=1:
M
addw=addw+w(i);
end
fori=1:
M
yi(i)=p0_1(i)+p1_1(i)*x
(1)+p2_1(i)*x
(2)+p3_1(i)*x(3)+p4_1(i)*x(4)+p5_1(i)*x(5)+p6_1(i)*x(6);
end
addyw=0;
fori=1:
M
addyw=addyw+yi(i)*w(i);
end
%计算输出
szzb(k)=addyw/addw;
end
szzbz3=mapminmax('reverse',szzb,outputps);
fori=1:
m
ifszzbz3(i)<=1.5
szpj3(i)=1;
elseifszzbz3(i)>1.5&&szzbz3(i)<=2.5
szpj3(i)=2;
elseifszzbz3(i)>2.5&&szzbz3(i)<=3.5
szpj3(i)=3;
elseifszzbz3(i)>3.5&&szzbz3(i)<=4.5
szpj3(i)=4;
else
szpj3(i)=5;
end
end
figure(3)
plot(szzbz1,'o-r')
holdon
plot(szzbz2,'*-g')
holdon
plot(szzbz3,'*:
b')
xlabel('时间','fontsize',12)
ylabel('预测水质','fontsize',12)
legend('C','B','A','fontsize',12)
functionf=ex16_16(x)
%输入各个点的坐标值
a=[21593126201811];
b=[1091318135786];
f
(1)=abs(x
(1)-a
(1))+abs(x
(2)-b
(1));
f
(2)=abs(x
(1)-a
(2))+abs(x
(2)-b
(2));
f(3)=abs(x
(1)-a(3))+abs(x
(2)-b(3));
f(4)=abs(x
(1)-a(4))+abs(x
(2)-b(4));
f(5)=abs(x
(1)-a(5))+abs(x
(2)-b(5));
f(6)=abs(x
(1)-a(6))+abs(x
(2)-b(6));
f(7)=abs(x
(1)-a(7))+abs(x
(2)-b(7));
f(8)=abs(x
(1)-a(8))+abs(x
(2)-b(8));
f(9)=abs(x
(1)-a(9))+abs(x
(2)-b(9));
f(10)=abs(x
(1)-a(10))+abs(x
(2)-b(10));
- 配套讲稿:
如PPT文件的首页显示word图标,表示该PPT已包含配套word讲稿。双击word图标可打开word文档。
- 特殊限制:
部分文档作品中含有的国旗、国徽等图片,仅作为作品整体效果示例展示,禁止商用。设计者仅对作品中独创性部分享有著作权。
- 关 键 词:
- 神经网络 优化 学习 程序