竞争网络权值分析

落日映苍穹つ 2022-05-16 10:27 174阅读 0赞

竞争网络权值分析

%竞争网络权值分析 clc; clear; P=\[0.7071 0.6402 0.000 -0.1961 0.1961 -0.9285 -0.8762 -0.8192; 0.7071 0.7682 -1.000 -0.9806 -0.9806 0.3714 0.4819 0.5735\]; S=4; \[R,Q\]=size(P); net=newc(minmax(P),S); %创建竞争网络 % Create a competitive layer % % Competitive layers are used to solve classification problems. % net = newc creates a new network with a dialog box. % net = newc(PR,S,KLR,CLR) takes these inputs, % PR -- R x 2 matrix of min and max values for R input elements % S -- Number of neurons % KLR -- Kohonen learning rate, default = 0.01 % CLR -- Conscience learning rate, default = 0.001 % and returns a new competitive layer. net.iw\{1,1\}=randnr(S,2); %赋归一化后的权值 w1=net.iw\{1,1\} %输出权值 b1=w1'; %%%%%%%%画圆开始 x=0:0.1:3.1416\*2; plot(sin(x),cos(x)); %%%%%%%%画圆结束 axis('equal') xlabel('P(1,q)..+ W(i,1)..o') ylabel('P(2,q) W(i,2)') hold on plot(P(1,:),P(2,:),'r+') %画输入状态点 plot(b1(1,:),b1(2,:),'o') %画初始权值点 %%%%%%%%%%%%%画输入点和原点的连线 for i=1:Q plot(\[P(1,i) 0\],\[P(2,i) 0\]) end %%%%%%%%%%%%%画输入点和原点的连线 hold off % disp('按任一键继续 ') % pause %%%%%%%%%%%%%%%%%%网络训练开始 net.trainParam.epochs=320; %最大训练次数 net.trainParam.show=100; %显示间隔 lp.lr=0.05; net=train(net,P); %训练竞争网络 %%%%%%%%%%%%%%%%%%网络训练结束 w2=net.iw\{1,1\} %输出训练后的权值 b2=w2'; a=zeros(4,Q)+compet(net.iw\{1,1\}\*P) %输出训练后的结果 % compet is a transfer function. % Transfer functions calculate a layer's output from its net input. % compet(N) takes one input argument, % N - S x Q matrix of net input (column) vectors. % and returns output vectors with 1 % where each net input vector has its maximum value, % and 0 elsewhere. % n = \[0; 1; -0.5; 0.5\]; % a = compet(n); % subplot(2,1,1), plot(n), ylabel('n') % subplot(2,1,2), plot(a), ylabel('a') % a = % % (2,1) 1 x=0:0.1:3.1416\*2; figure(2) plot(sin(x),cos(x)); %画圆 axis('equal') xlabel('P(1,q)..+ W(i,1)..o') ylabel('P(2,q) W(i,2)') hold on plot(P(1,:),P(2,:),'r+') plot(b2(1,:),b2(2,:),'o') %作训练后的权值点 for i=1:Q plot(\[P(1,i) 0\],\[P(2,i) 0\]) end hold off

输出如下：

w1 = -0.7318 -0.6815 -0.8692 0.4944 -0.1281 -0.9918 0.4407 -0.8977 TRAINR, Epoch 0/320 TRAINR, Epoch 100/320 TRAINR, Epoch 200/320 TRAINR, Epoch 300/320 TRAINR, Epoch 320/320 TRAINR, Maximum epoch reached. w2 = -0.0671 0.6655 -0.8995 0.4316 -0.0892 -0.9900 0.4553 -0.1103 a = 1 1 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0