source: distools/disnorm.m

%DISNORM Normalization of a dissimilarity matrix
%
%       V = DISNORM(D,OPT)
%   F = E*V
%
% INPUT
%   D    NxN dissimilarity matrix or dataset, which sets the norm
%   E    Matrix to be normalized, e.g. D itself
%       OPT  'max' : maximum dissimilarity is set to 1 by global rescaling
%                  'mean': average dissimilarity is set to 1 by global rescaling (default)
%
% OUTPUT
%   V   Fixed mapping
%   F   Normalized dissimilarity data
%
% DEFAULT
%   OPT = 'mean'
%
% DESCRIPTION
% Operation on dissimilarity matrices, like the computation of classifiers
% in dissimilarity space, may depend on the scaling of the dissimilarities
% (a single scalar for the entire matrix). This routine computes a scaling
% for a giving matrix, e.g. a training set and applies it to other
% matrices, e.g. the same training set or based on a test set.

% Copyright: Elzbieta Pekalska, ela.pekalska@googlemail.com
% Faculty EWI, Delft University of Technology and
% School of Computer Science, University of Manchester


function V = disnorm(D,opt)
if nargin < 2,
        opt = 'mean';
end

if nargin == 0 | isempty(D)
  V = prmapping(mfilename,{opt});
  V = setname(V,'Disnorm');
  return
end

if ~isdataset(D)
        D = prdataset(D,1);
        D = setfeatlab(D,getlabels(D));
end

%DEFINE mapping
if isstr(opt)
%       discheck(D);
        opt = lower(opt);
        if strcmp(opt,'mean')
                n = size(D,1);
                m = sum(sum(+D))/(n*(n-1));
        elseif strcmp(opt,'max')
                m = max(D(:));
        else
                error('Wrong OPT.')
        end
        if nargout > 1
                D = D./m;
        end     
    V = prmapping(mfilename,'trained',{m},[],size(D,2),size(D,2));
        return;
end


% APPLY mapping
if ismapping(opt)
        opt = getdata(opt,1);
        V = D./opt;
        return;
end