corproen.m

FISMAT accommodates different arithmetic operators, fuzzification and defuzzification algorithm, imp

function O=corproen(FAMbank,BeA,BdeA,L)

% O=corproen(FAMbank,BeA,BdeA,L)  
%
% Correlation-Product Encoding. Kosko (1992)
%
% Correlation-product encoding :
% Defining pointwise the fuzzy Hebb matrix hebbbM by the minimum of the 
% "signals" BdeA(i) and BeA(j), an encoding scheme we shall call 
% correlation-minimum encoding. Fuzzy correlation constructs fuzzy matrices. 
% Correlation-minimum encoding produces a matrix of clipped BeA sets, while
% the correlation-product encoding produces a matrix of scaled BeA sets. 
% In practise we sidestep this large numerical matrix with a virtual 
% representation scheme. In place of the matrix the user encodes the fuzzy-
% set associations as linguistic entries in a FAM-bank linguistic matrix.
%
% L is the library of fuzzy sets for the output universe of discourse Y.
% The universe of discourse is the most right collumn of L.
% The system sums all O(i) to form the combined output fuzzy set O.
% O is a row vector.
%
% FSTB - Fuzzy Systems Toolbox for MATLAB
% Copyright (c) 1993-1996 by Olaf Wolkenhauer
% Control Systems Centre at UMIST
% Manchester M60 1QD, UK
%
% 28-Feb-1994
 
% BeA and BdeA are the product of the bit vector B for each input and the 
% matrix A which contains the library of fuzzy sets.
% Correllation minimum encoding => fuzzy Hebb matrix :

A=BdeA';
n=length(A); p=length(BeA); hebbM=zeros(n,p);
for i=1:n,
  hebbM(i,:)=min((ones(1,p).*A(i)),BeA);
end;

% The hebb matrix contains now the w(i), which determines the degree, to 
% which the corresponding rule activates the consequent fuzzy set. 
% For the correllation-minimum encoding we, we clip the consequent fuzzy set 
% L(i) in the library of fuzzy-set values to the w(i) degree with pointwise
% minimum. mOi(y)=min(w(i),mLi(Y)).

% The library L is already rotated (rot90() and flipud()). 
% That means in each collumn one set. The most left represents Large Neg.
% From mLi(ymin) at the top to mLi(ymax) at the bottom. 

[n,N]=size(L);                 % n=dimension of L. N number of library sets.
[i,j]=find(FAMbank);           % "coordinates" of existing rules.
Oi=ones(length(i),n);
for r=1:length(i),
  Oi(r,:)=hebbM(i(r),j(r)).*Oi(r,:).*(L(:,FAMbank(i(r),j(r))))';
end;
O=ones(1,length(i))*Oi; % O is a row vector with n elements.