scale.m

% scale() -  Scaling the data for libsvm 
%
% Usage:
%    >> scale(x)
%
% Inputs:
%   x               - An m by n matrix of m training instances with n features
% Outputs:  
%   y               - Scaled data , every data divided by max value of
%                      this column
%   maxvalue        - Max value of every column

function [scaled,MaxV,MinV] = scale(Data, Lower, Upper)
% [scaled] = Scale(Data);
% [scaled] = Scale(Data, Lower, Upper);
% 
% scale the elements of all the column vectors in
% the range of [Lower Upper]. (default [-1 1])

if (nargin<3) 
%    Lower = -1;
%    Upper = 1;
  Lower = 0;
   Upper = 1;
elseif (Lower > Upper)
   disp (['Wrong Lower or Upper values!']);
end

[MaxV, I]=max(Data);
[MinV, I]=min(Data);

[R,C]= size(Data);

scaled=(Data-ones(R,1)*MinV).*(ones(R,1)*((Upper-Lower)*ones(1,C)./(MaxV-MinV)))+Lower;