%   CmpLrn.m
%            A.P. Paplinski   27 August 2004
%   A simple competitive learning 

figure(1), clf reset

CmpInit
W = Win ;
Whist = zeros( N, m, p);  % to store all weights
Whist(1,:,:) = W ;  
wnm = ones(m,1) ;

eta = 0.1 ;     % learning gain
% deta = 1-1/N ;  % learning gain decaying factor

for  n = 1:N
  xn = X(:,n)' ;
% the current vector is compared with all weight vectors
  xmw = xn(wnm,:)-W ;
  [win kwin] = min(sum((xmw.^2),2)) ;
% the weights of the winning neurons are update
  W(kwin,:) = W(kwin,:) + eta*xmw(kwin,:) ;
  Whist(n,:,:) = W ;
% eta = eta*deta ;
% end 

plot(X(1,:),X(2,:),'g.',clst(1,:),clst(2,:),'ro', ...
     Win(:,1),Win(:,2),'b^' , ...
     Whist(1:n,:,1), Whist(1:n,:,2), 'b', W(:,1),W(:,2) , 'r*'),  
  title(['Simple Competitive Learning    n = ', num2str(n)])
  xlabel('x_1'), ylabel('x_2'), grid on
%  pause(1)
end

%  myfigA(1, 10,9)
plot(X(1,:),X(2,:),'g.',clst(1,:),clst(2,:),'ro', ...
     Win(:,1),Win(:,2),'b^' , ...
     Whist(1:n,:,1), Whist(1:n,:,2), 'b', W(:,1),W(:,2) , 'r*'), grid
  title(['Simple Competitive Learning    n = ', num2str(n)])
 xlabel('x_1'), ylabel('x_2')
% print -f1 -depsc2 CmpLrn