function predicted = cosmo_classify_matlabsvm_2class(samples_train, targets_train, samples_test, opt)
% svm classifier wrapper (around svmtrain/svmclassify)
%
% predicted=cosmo_classify_matlabsvm_2class(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 struct with options. supports any option that
% svmtrain supports
%
% Output:
% predicted Qx1 predicted data classes for samples_test
%
% Notes:
% - this function uses Matlab's builtin svmtrain function, which has
% the same name as LIBSVM's version. Use of this function is not
% supported when LIBSVM's svmtrain precedes in the matlab path; in
% that case, adjust the path or use cosmo_classify_libsvm instead.
% - 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.
% - for a guide on svm classification, see
% http://www.csie.ntu.edu.tw/~cjlin/papers/guide/guide.pdf
% Note that cosmo_crossvalidate and cosmo_crossvalidation_measure
% provide an option 'normalization' to perform data scaling
% - As of Matlab 2017a (maybe earlier), Matlab gives the warning that
% 'svmtrain will be removed in a future release. Use fitcsvm instead.'
% however fitcsvm gives different results than svmtrain; as a result
% cosmo_classify_matlabcsvm gives different results than
% cosmo_classify_matlabsvm. In this function the warning message is
% . suppressed.
% - As of Matlab 2018a, this function cannot be used anymore. Use
% cosmo_classify_matlabcsvm instead.
%
% See also svmtrain, svmclassify, cosmo_classify_matlabsvm,
% cosmo_classify_matlabcsvm
%
% # For CoSMoMVPA's copyright information and license terms, #
% # see the COPYING file distributed with CoSMoMVPA. #
if nargin < 4
opt = struct();
end
[ntrain, nfeatures] = size(samples_train);
[ntest, nfeatures_] = size(samples_test);
ntrain_ = numel(targets_train);
if nfeatures ~= nfeatures_ || ntrain_ ~= ntrain
error('illegal input size');
end
if nfeatures == 0
% matlab's svm cannot deal with empty data, so predict all
% test samples as the class of the first sample
predicted = targets_train(1) * ones(ntest, 1);
return
end
classes = unique(targets_train);
nclasses = numel(classes);
if nclasses ~= 2
error(['%s requires 2 classes, found %d. Consider using '...
'cosmo_classify_{matlab,lib}svm instead'], ...
mfilename(), nclasses);
end
opt_cell = opt2cell(opt);
% train & test; if it fails, see if this caused by non-functioning
% matlabsvm
try
if ~cached_has_matlabsvm()
cosmo_check_external('matlabsvm');
end
% Use svmtrain and svmclassify to get predictions for the testing set.
% disable warnings shown by Matlab 2017 and later
orig_warning_state = warning();
cleaner = onCleanup(@()warning(orig_warning_state));
warning('off', 'stats:obsolete:ReplaceThisWith');
warning('off', ['stats:obsolete:ReplaceThisWith'...
'MethodOfObjectReturnedBy']);
% store most recent warning
[orig_lastmsg, orig_lastid] = lastwarn();
model = svmtrain(samples_train, targets_train, opt_cell{:});
predicted = svmclassify(model, samples_test);
% deal with possible warning shown (Matlab >= 2016)
[warning_msg, warning_id] = lastwarn();
if strcmp(warning_id, 'stats:obsolete:ReplaceThisWith')
% only show warning once (by default) if this is a
% a stats:obsolete message
suffix = ['CoSMoMVPA note: the more recent '...
'fitcsvm / svmsmoset classifiers produce '...
'different results '...
'than the older svmtrain function. '...
'To use fitcsvm, use cosmo_classify_matlabcsvm'];
cosmo_warning('%s\n%s', warning_msg, suffix);
elseif ~strcmp(warning_id, orig_lastid)
% new warning was issued , different from stats:obsolete one;
% show warning message
cosmo_warning(warning_id, warning_msg);
end
catch
caught_exception = lasterror();
cosmo_check_external('matlabsvm');
if strcmp(caught_exception.identifier, ...
'stats:svmtrain:NoConvergence')
error(['SVM training did not converge. Your options are:\n'...
' 1) increase ''boxconstraint''\n'...
' 2) increase ''tolkkt''\n'...
' 3) set ''kktviolationlevel'' to a positive value\n'...
' 4) use a different classifier\n'...
'If you do not have a strong preference for '...
'either option, you are advised to try option (4) '...
'using cosmo_classify_lda'], '');
else
rethrow(caught_exception);
end
end
% helper function to convert cell to struct
function opt_cell = opt2cell(opt)
if isempty(opt)
opt_cell = cell(0);
return
end
to_keep = {'kernel_function', ...
'rbf_sigma', ...
'polyorder', ...
'mlp_params', ...
'method', ...
'options', ...
'tolkkt', ...
'kktviolationlevel', ...
'kernelcachelimit', ...
'boxconstraint', ...
'autoscale', ...
'showplot'};
fns = fieldnames(opt);
keep_msk = cosmo_match(fns, to_keep);
keep_fns = fns(keep_msk);
keep_id = find(keep_msk);
n = numel(keep_fns);
opt_cell = cell(1, 2 * n);
for k = 1:n
fn = fns{keep_id(k)};
opt_cell{k * 2 - 1} = fn;
opt_cell{k * 2} = opt.(fn);
end
function tf = cached_has_matlabsvm()
persistent cached_tf
if isequal(cached_tf, true)
tf = true;
return
end
cached_tf = cosmo_check_external('matlabsvm');
tf = cached_tf;
