ojaalgorithm.m

內涵模糊理論與類神經網路的程式碼...提供初學者做研究參考

% ==========================================================
% 
%           Neural Networks A Classroom Approach
%                     Satish Kumar
%             Copyright Tata McGraw Hill, 2004
%
%        MATLAB code that implements Oja's algorithm 
%        for a linear neuron
%             Reference: Table 12.2;Page 530
%
% ==========================================================

% MATLAB Program to simulate Oja's Rule

x=0.5*randn(500,1);		% Generate X scatter
y=0.05*randn(500,1);		% Generate Y scatter
input=[x';y'];				% Create input data matrix
theta = 0;
r=[cos(theta) -sin(theta)
   sin(theta) cos(theta)];
inputnew=r*input;

xshift=0;
yshift=0;

for i=1:500
   inputnew(1,i)=inputnew(1,i)+xshift;
   inputnew(2,i)=inputnew(2,i)+yshift;
end

figure
hold on
zoom on
plot(inputnew(1,:),inputnew(2,:),'.k');	% Plot scatter
axis equal
grid on
eta=0.15;					% Initialize learning rate
w=[.1;.5];					% and the weights
plot(0.1,0.5,'o');		% Mark the initial point
text(0.08,0.6,'\downarrow \fontname{times} Initial weight (0.1,0.5)')

for epoch=1:15				% We'll do 15 epochs
   for i=1:500
      s = inputnew(:,i)' * w;	% Compute activation
      w = w + eta * s * (input(:,i) - s*w);	% Update weights
      plot(w(1),w(2),'.','markersize',10);	% Plot weight point
   end
end

plot(w(1),w(2),'o');
axis([-2 2 -0.5 1]);
text(1,-0.15,'\uparrow \fontname{times} Final weight');
text(1,-0.3,'(1.0,-.004)');

xlabel('x');
ylabel('y');
title('\itLinear neuron simulation on 2d normal data using Ojas rule');