wolffd@0: function d=som_eucdist2(Data, Proto)
wolffd@0: 
wolffd@0: %SOM_EUCDIST2 Calculates matrix of squared euclidean distances between set of vectors or map, data struct
wolffd@0: %
wolffd@0: % d=som_eucdist2(D, P)
wolffd@0: %
wolffd@0: %  d=som_eucdist(sMap, sData);
wolffd@0: %  d=som_eucdist(sData, sMap);
wolffd@0: %  d=som_eucdist(sMap1, sMap2);
wolffd@0: %  d=som_eucdist(datamatrix1, datamatrix2);
wolffd@0: %
wolffd@0: %  Input and output arguments ([]'s are optional): 
wolffd@0: %   D (matrix) size Nxd
wolffd@0: %     (struct) map or data struct
wolffd@0: %   P (matrix) size Pxd
wolffd@0: %     (struct) map or data struct
wolffd@0: %   d (matrix) distance matrix of size NxP 
wolffd@0: %
wolffd@0: % IMPORTANT
wolffd@0: %
wolffd@0: % * Calculates _squared_ euclidean distances
wolffd@0: % * Observe that the mask in the map struct is not taken into account while 
wolffd@0: %   calculating the euclidean distance
wolffd@0: %
wolffd@0: % See also KNN, PDIST.
wolffd@0: 
wolffd@0: % Contributed to SOM Toolbox 2.0, October 29th, 2000 by Johan Himberg
wolffd@0: % Copyright (c) by Johan Himberg
wolffd@0: % http://www.cis.hut.fi/projects/somtoolbox/
wolffd@0: 
wolffd@0: % Version 2.0beta Johan 291000
wolffd@0: 
wolffd@0: %% Init %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
wolffd@0: 
wolffd@0: if isstruct(Data);
wolffd@0:   if isfield(Data,'type') & ischar(Data.type),
wolffd@0:     ;
wolffd@0:   else
wolffd@0:     error('Invalid map/data struct?');
wolffd@0:   end
wolffd@0:   switch Data.type
wolffd@0:    case 'som_map'
wolffd@0:     data=Data.codebook;
wolffd@0:    case 'som_data'
wolffd@0:     data=Data.data;
wolffd@0:   end
wolffd@0: else
wolffd@0:   % is already a matrix
wolffd@0:   data=Data;
wolffd@0: end
wolffd@0: 
wolffd@0: % Take prototype vectors from prototype struct
wolffd@0: 
wolffd@0: if isstruct(Proto),
wolffd@0:   
wolffd@0:   if isfield(Proto,'type') & ischar(Proto.type),
wolffd@0:     ;
wolffd@0:   else
wolffd@0:     error('Invalid map/data struct?');
wolffd@0:   end
wolffd@0:   switch Proto.type
wolffd@0:    case 'som_map'
wolffd@0:     proto=Proto.codebook;
wolffd@0:    case 'som_data'
wolffd@0:     proto=Proto.data;
wolffd@0:   end
wolffd@0: else
wolffd@0:   % is already a matrix
wolffd@0:   proto=Proto; 
wolffd@0: end
wolffd@0: 
wolffd@0: % Check that inputs are matrices
wolffd@0: if ~vis_valuetype(proto,{'nxm'}) | ~vis_valuetype(data,{'nxm'}),
wolffd@0:   error('Prototype or data input not valid.')
wolffd@0: end
wolffd@0: 
wolffd@0: % Record data&proto sizes and check their dims 
wolffd@0: [N_data dim_data]=size(data); 
wolffd@0: [N_proto dim_proto]=size(proto);
wolffd@0: if dim_proto ~= dim_data,
wolffd@0:   error('Data and prototype vector dimension does not match.');
wolffd@0: end
wolffd@0: 
wolffd@0: % Calculate euclidean distances between classifiees and prototypes
wolffd@0: d=distance(data,proto);
wolffd@0: 
wolffd@0: %%%% Classification %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
wolffd@0: function d=distance(X,Y);
wolffd@0: 
wolffd@0: % Euclidean distance matrix between row vectors in X and Y
wolffd@0: 
wolffd@0: U=~isnan(Y); Y(~U)=0;
wolffd@0: V=~isnan(X); X(~V)=0;
wolffd@0: d=abs(X.^2*U'+V*Y'.^2-2*X*Y');