prototype.py

train_feats = sg.RealSparseFeatures('../train.dat')
test_feats = sg.RealSparseFeatures('../test.dat')
train_labels = sg.RealDenseLabels('../train_labels.dat')

gammas = np.arange(1, 10)

lda = sg.LDA()
lda.preprocessor_chain().enque(sg.DimensionSubset()).enque(sg.NormOne()).enque(sg.FisherLDA())

params = sg.ModelSelectionParameters()
params.set_values(gammas, sg.MSPT_INT32)

splitting_strategy = StratifiedCrossValidationSplittingStrategy()

cv = CrossValidation(lda, train_feats, train_labels, splitting_strategy, MeanSquarredError()).parallel()

ms = sg.GridSearchModelSelection(cv, params)
param = ms.select_model().get_parameter()

lda.set_gamma(param)
lda.set_features(train_feats)
lda.set_labels(train_labels)

lda.train()

predicted_labels = lda.apply_binary(test_feats)