rankboostauc.m

应用于matlab支持向量机svm的工具箱

function [alpha,threshold,rankfeat]=rankboostAUC(xapp,yapp,T);

%  USAGE
%
% [alpha,threshold,rankfeat]=rankboostAUC(xapp,yapp,T);
%
%  This a Rankboost algorithm as described in the freund et al
%  Journal of Machine Learning Research paper.
%  
%  xapp and yapp are the learning set data and labels
%  T is the number of weak learners which is a step function.
%
%  the outputs are 
%  
%   alpha  : vector  of weigth of each weak learner
%   threshold : vector of each weak learner translation
%   rankfeat : vector of each weak learner feature.
%
%  see also rankboostAUCval

% 30/07/2004  A. Rakotomamonjy
%
%



[n,dim]=size(xapp);
indpos=find(yapp==1);
npos=length(indpos);
indneg=find(yapp==-1);
nneg=length(indneg);
yapp=[yapp(indpos);yapp(indneg)];    % on remet les donn閑s en ordre!
xapp=[xapp(indpos,:);xapp(indneg,:)];
indpos=find(yapp==1);
indneg=find(yapp==-1);

% dans les vecteurs on stocke d'abord les v(xo) puis les v(x1) xo c'est les
% pos et x1 les neg par rapport