ANN.rar_self organizing map_神经网络 排序_网络映射 matlab_自组织映射

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%建立神经网络%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% clc clear % 设定样本数据 m=9; n=8; P=rand(m,n); for i=1:m for j=1:n if P(i,j)>0.7 P(i,j)=1; else P(i,j)=0; end end end %转置后符合神经网络的输入格式 P=P'; %newsom建立SOM网络,竞争层为6*7=42个神经元 net=newsom(minmax(P),[6 7]); plotsom(net.layers{1}.positions) %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%建立好神经网络后，分别设定8种训练的步数%%%%%%%%% %8次训练的步数 a=[5 20 40 80 160 320 640 1280]; % 随机初始化一个向量 yc=rands(9,9); % 训练步数为5 net.trainparam.epochs=a(1); %训练网络和查看分类 net=train(net,P); %仿真网络 y=sim(net,P); yc(1,:)=vec2ind(y); figure(1) subplot(2,2,1); plotsom(net.IW{1,1},net.layers{1}.distances) % 训练步数为20 net.trainparam.epochs=a(2); % 训练网络和查看分类 net=train(net,P); %仿真网络 y=sim(net,P); yc(2,:)=vec2ind(y); subplot(2,2,2); plotsom(net.IW{1,1},net.layers{1}.distances) % 训练步数为40 net.trainparam.epochs=a(3); % 训练网络和查看分类 net=train(net,P); %仿真网络 y=sim(net,P); yc(3,:)=vec2ind(y); subplot(2,2,3); plotsom(net.IW{1,1},net.layers{1}.distances) % 训练步数为80 net.trainparam.epochs=a(4); % 训练网络和查看分类 net=train(net,P); %仿真网络 y=sim(net,P); yc(4,:)=vec2ind(y); subplot(2,2,4); plotsom(net.IW{1,1},net.layers{1}.distances) % 训练步数为160 net.trainparam.epochs=a(5); % 训练网络和查看分类 net=train(net,P); %仿真网络 y=sim(net,P); yc(5,:)=vec2ind(y); subplot(2,2,1); plotsom(net.IW{1,1},net.layers{1}.distances) % 训练步数为320 net.trainparam.epochs=a(6); % 训练网络和查看分类 net=train(net,P); %仿真网络 y=sim(net,P); yc(6,:)=vec2ind(y); subplot(2,2,2); plotsom(net.IW{1,1},net.layers{1}.distances) % 训练步数为640 net.trainparam.epochs=a(7); % 训练网络和查看分类 net=train(net,P); %仿真网络 y=sim(net,P); yc(7,:)=vec2ind(y); subplot(2,2,3); plotsom(net.IW{1,1},net.layers{1}.distances) % 训练步数为1280 net.trainparam.epochs=a(8); % 训练网络和查看分类 net=train(net,P); %仿真网络 y=sim(net,P); yc(8,:)=vec2ind(y); subplot(2,2,4); plotsom(net.IW{1,1},net.layers{1}.distances)