art.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 the ART 1 algorithm
%             Reference: Table 11.4;Page 505
%
% ==========================================================

% ART 1 ALGORITHM FOR BINARY PATTERNS
% Classify the following four 5-dimensional vectors
% Use L=2, and set the vigilance as desired.

n = 5;            % Initialize n,m, and rho
m = 1;
rho = 0.3;
out = ones(m,n);  % Outstar initalized to 1
inwt = 2/(1+n);   % Initial value of each instar weight
in = (inwt*out)'; % Instar set up

p=[1 1 1 0 0		% Store the patterns
1 1 0 0 0 
0 0 0 0 1
0 0 0 1 1];
q = 4;            % q patterns to be classified

newnodeflag = 1;  % Get into the loop 

while(newnodeflag==1)      % New node added to F2...
  newnodeflag =0;          % reset the newnode flag...
  for k = 1:q              % for each pattern
    outstarlearn = 0       % reset the outstar learnt flag
    index = ones(1,m)      % all nodes can compete 
    y = p(k,:)*in	   		% compute F2 activations 
    while(sum(index) ~= 0) % some nodes not reset
      [maxy,windex] = max(y)    % find index of winner
      s = out(windex,:).*p(k,:) % compute F1 signals
      M = sum(s)/sum(p(k,:))    % Find ratio of signals, M
      if M > rho                % check ratio with vigilance
        out(windex,:) = s       % if ok then learn
        in(:,windex) = (2/(1+sum(s)))*s'
        outstarlearn = 1        % set learnt flag
        break                   % exit
      else
        index(windex) = 0       % else reset the index entry
        y(windex) = -1          % suppress activity of M
      end                       % (they can never be negative!)
    end
    
    % No node could classify
    if ((~outstarlearn) & sum(index) == 0)
         m = m+1          % add a new node
         out(m,:)=p(k,:)  % Learn
         in(:,m)=p(k,:)'*(2/(1+sum(p(k,:))))  
         newnodeflag = 1  % set the new node added flag
    end
 end
end