Nearest-Neighbour-Search-Neural-Networks-Product-Quantization/TEXMEX/PQ Library/Quantization.m

%% Clusters Distance
%  The function provides the computation of Euclidean Distance
%  Between Clusters and input Data Matrices.
%  If the flag CMinDistON is active, it returns the Cluster Matrix
%  corresponding to input Data which are minimum at distance.
% 
%  Inputs: 
% - XSet        Input Matrix: each row is a multidimensional data
% - CSet        Cluster Matrix: each row is a multidimensional data
% - CMinDistON  Flag which allows to return the Minimum Distance Clusters
% 
%  Outputs:
% - XCDist      Matrix containing the Distances between Data and Clusters
% - CMinDist    Matrix containing the Clusters at Minimum Distance 
%%

function [ CMinDist, iCMinDist] = Quantization( XSet, CSet)

%     Ntr = size(XSet,2);
%     imgSzSq=size(CSet,1);
%     kCent = size(CSet,2);
%     
%     XCDist=zeros(kCent,Ntr);
%     CMinDist=zeros(imgSzSq,Ntr);
%     iCMinDist=zeros(1,Ntr);
%     
%     for jj=1:Ntr
%         for ii=1:kCent
%             XCDist(ii,jj)=sqrt(sum((XSet(:,jj)-CSet(:,ii)).^2)); 
%         end
%             [~,iCMinDist(jj)]=min(XCDist(:,jj));
%             CMinDist(:,jj)=CSet(:,iCMinDist(jj));
%     end
    
    
  CSetSq = sum (CSet.^2) / 2;
  XSetSq = sum (XSet.^2) / 2;

  XCDist = bsxfun (@plus, XSetSq', bsxfun (@minus, CSetSq, XSet'*CSet));

%   if nRecall == 1
    [CMinDist, iCMinDist] = min (XCDist, [], 2);
%   else  
%     [CMinDist, iCMinDist] = sort (XCDist, 2);
%     CMinDist = CMinDist (:, 1:nRecall);
%     iCMinDist = iCMinDist (:, 1:nRecall);
%   end
  
%  CMinDist = CMinDist' * 2;
%  iCMinDist = iCMinDist';
end