%FEATEVAL Evaluation of feature set for classification
%
% J = feateval_new(A,CRIT,T)
% J = feateval_new(A,CRIT,N)
%
% INPUT
% A input dataset
% CRIT string name of a method or untrained mapping
% T validation dataset (optional)
% N number of cross-validations (optional)
%
% OUTPUT
% J scalar criterion value
%
% DESCRIPTION
% Evaluation of features by the criterion CRIT for classification,
% using objects in the dataset A. The larger J, the better. Resulting
% J-values are incomparable over the various methods.
% The following methods are supported:
%
% crit='in-in' : inter-intra distance.
% crit='maha-s': sum of estimated Mahalanobis distances.
% crit='maha-m': minimum of estimated Mahalanobis distances.
% crit='eucl-s': sum of squared Euclidean distances.
% crit='eucl-m': minimum of squared Euclidean distances.
% crit='NN' : 1-Nearest Neighbour leave-one-out
% classification performance (default).
% (performance = 1 - error).
%
% CRIT can also be any untrained classifier, e.g. LDC([],1e-6,1e-6).
% The classification error is used for a performance estimate. If
% supplied, the dataset T is used for obtaining an unbiased estimate
% of the performance of classifiers trained with the dataset A.
% If a number of cross-validations N is supplied, the routine is
% run for N times with different training and test sets generated
% from A by cross-validation. Results are averaged. If T nor N are
% given, the apparent performance on A is used.
%
% SEE ALSO
% DATASETS, FEATSELO, FEATSELB, FEATSELF, FEATSELP, FEATSELM, FEATRANK
% Copyright: R.P.W. Duin, duin@ph.tn.tudelft.nl
% Faculty of Applied Physics, Delft University of Technology
% P.O. Box 5046, 2600 GA Delft, The Netherlands
% REVISIONS
% DXD1: David Tax, 08-05-2003
% I added the inter/intra distance criterion.
% $Id: feateval.m,v 1.8 2005/12/15 20:43:25 duin Exp $
function J = feateval_new(a,crit,t, varargin)
prtrace(mfilename);
[ma,k,c] = getsize(a);
if nargin < 2
crit = 'NN';
end
if nargin < 3
t =[];
prwarning(4,'Where needed, input dataset is used for validation')
end
if(isscalar(t) && ~isdataset(t) ) % cross-validation desired, t rotations
if( t > 0 )
K = crossval(a,nmc,t,0); % trick to get rotation set from crossval
J = 0;
JALL = 1:size(a,1);
for j=1:t
JIN = JALL;
JOUT = find(K==j);
JIN(JOUT) = [];
JOUT = JALL(JOUT);
train = a(JIN,:);
test = a(JOUT,:);
J = J + feval(mfilename,train,crit,test);
end
J = J/t;
return
else
J = cross_val_leave1out_rec_index(a, crit);
return;
end
end
% if(~isempty(t))
% ClassSizes = classsizes(t);
% bClassesPresent_t = ClassSizes > 0 ;
% else
% ClassSizes = classsizes(a);
% bClassesPresent_t = ClassSizes > 0 ;
% end
%
% if( sum(bClassesPresent_t) < 2 )
% error('Debe haber representadas al menos 2 clases en el dataset.');
% end
% islabtype(a,'crisp');
isvaldset(a,1,2); % at least 1 object per class, 2 classes
iscomdset(a,t);
if ischar(crit)
%DXD1
if strcmp(crit,'in-in') % inter/intra distances
islabtype(a,'crisp','soft');
if isempty(t)
[U,G] = meancov_new(a,0);
prior = getprior(a);
else
[U,G] = meancov_new(t,0);
prior = 1/sum(classsizes(t))*classsizes(t);
% S_w = sum(G(:,:,bClassesPresent_t),3);
%
% iClassMean = +U;
% iTotalMean = prior(bClassesPresent_t) * iClassMean(bClassesPresent_t,:);
%
% S_b = zeros( size(iClassMean,2) );
%
% for j = find(bClassesPresent_t)
% iAux = iClassMean(j,:) - iTotalMean; % between scatter
% S_b = S_b + prior(j)*(iAux' * iAux);
% end
end
S_w = reshape(nansum(reshape(G(:,:,bClassesPresent_t),k*k,sum(bClassesPresent_t))*prior(bClassesPresent_t)',2),k,k); % within scatter
[dummy, S_b] = meancov_new(+U, 0); % between scatter
if( det(S_w) == 0 )
J = 0;
else
J = trace(inv(S_w)*S_b);
end
% t
% U
% +U
% disp(['G:' num2str(G) ' Sw:' num2str(S_w) ' S_b:' num2str(S_b)]);
elseif( strcmp(crit,'maha-s') || strcmp(crit,'maha-m') )% Mahalanobis distances
islabtype(a,'crisp','soft');
if isempty(t)
D = distmaha_new(a,[],[]);
else
[U,G] = meancov_new( a, 0 );
D = distmaha(t,U,G);
D = meancov_new(D,0);
end
D = +D;
D = D(bClassesPresent_t,bClassesPresent_t);
if strcmp(crit,'maha-m')
D = D + realmax*eye(sum(bClassesPresent_t));
J = min(min(D));
else
J = sum(sum(D))/2;
end
elseif strcmp(crit,'eucl-s') || strcmp(crit,'eucl-m') % Euclidean distances
islabtype(a,'crisp','soft');
U = meancov_new(a,0);
if(isempty(t))
D = distm_new(U,U);
else
D = distm_new(t,U);
D = meancov_new(D,0);
end
D = +D;
D = D(bClassesPresent_t,bClassesPresent_t);
if strcmp(crit,'eucl-m')
D = D + realmax*eye( sum(bClassesPresent_t) );
J = min(min(D));
else
J = sum(sum(D))/2;
end
elseif strcmp(crit,'NN') % 1-NN performance
islabtype(a,'crisp','soft');
if isempty(t)
J = 1 - testk(a,1);
else
J = 1 - testk(a,1,t);
end
else
error('Criterion undefined');
end
else
ismapping(crit);
isuntrained(crit);
if isempty(t)
J = 1 - (a * (a * crit) * testc);
else
J = 1 - (t * (a * crit) * testc);
end
end
return