function predicted=cosmo_classify_svm(samples_train, targets_train, samples_test, opt)
% classifier wrapper that uses either matlab's or libsvm's SVM.
%
% predicted=cosmo_classify_svm(samples_train, targets_train, samples_test, opt)
%
% Inputs:
% samples_train PxR training data for P samples and R features
% targets_train Px1 training data classes
% samples_test QxR test data
% opt (optional) struct with options for classification
% If a field 'svm' is present it should be either
% 'libsvm' or 'matlabsvm' to use that SVM. If this
% field is absent it will be selected automatically.
%
% Output
% predicted Qx1 predicted data classes for samples_test
%
% Notes:
% - cosmo_classify_svm can use either libsvm or matlab's svm, whichever is
% present
% - if both are present, then there is a conflict because 'svmtrain' is
% implemented differently by libsvm or matlab's svm. The path setting
% determines which svm implementation is used.
% - when using libsvm it requires version 3.18 or later:
% https://github.com/cjlin1/libsvm
% - for a guide on svm classification, see
% http://www.csie.ntu.edu.tw/~cjlin/papers/guide/guide.pdf
% - Matlab's SVM classifier is rather slow, especially for multi-class
% data (more than two classes). When classification takes a long time,
% consider using libsvm.
% - In both implemenations, by default the data is scaled.
% Note that cosmo_crossvalidate and cosmo_crossvalidation_measure
% provide an option 'normalization' to perform data scaling.
%
%
% See also: svmtrain, svmclassify, cosmo_classify_svm,
% cosmo_classify_libsvm, cosmo_crossvalidate,
% cosmo_crossvalidation_measure
%
% # For CoSMoMVPA's copyright information and license terms, #
% # see the COPYING file distributed with CoSMoMVPA. #
