import unittest
import sys, platform
from skmultilearn.adapt import MLkNN
from skmultilearn.cluster import LabelCooccurrenceGraphBuilder
from skmultilearn.embedding import CLEMS, SKLearnEmbedder, EmbeddingClassifier
from skmultilearn.tests.classifier_basetest import ClassifierBaseTest
from sklearn.linear_model import LinearRegression
from sklearn.manifold import SpectralEmbedding
from copy import copy
import sklearn.metrics as metrics
if not (sys.version_info[0] == 2 or platform.architecture()[0]=='32bit'):
from skmultilearn.embedding import OpenNetworkEmbedder
class EmbeddingTest(ClassifierBaseTest):
TEST_NEIGHBORS = 3
def classifiers(self):
graph_builder = LabelCooccurrenceGraphBuilder(weighted=True, include_self_edges=False)
param_dicts = {
'GraphFactorization': dict(epoch=1),
'GraRep': dict(Kstep=2),
'HOPE': dict(),
'LaplacianEigenmaps': dict(),
'LINE': dict(epoch=1, order=1),
'LLE': dict(),
}
if not (sys.version_info[0] == 2 or platform.architecture()[0] == '32bit'):
for embedding in OpenNetworkEmbedder._EMBEDDINGS:
if embedding == 'LLE':
dimension = 3
else:
dimension = 4
yield EmbeddingClassifier(
OpenNetworkEmbedder(copy(graph_builder), embedding, dimension,
'add', True, param_dicts[embedding]),
LinearRegression(),
MLkNN(k=2)
)
yield EmbeddingClassifier(
SKLearnEmbedder(SpectralEmbedding(n_components=2)),
LinearRegression(),
MLkNN(k=2)
)
EmbeddingClassifier(
CLEMS(metrics.accuracy_score, True),
LinearRegression(),
MLkNN(k=2),
True
)
def test_if_embedding_classification_works_on_sparse_input(self):
for classifier in self.classifiers():
self.assertClassifierWorksWithSparsity(classifier, 'sparse')
self.assertClassifierPredictsProbabilities(classifier, 'sparse')
def test_if_embedding_classification_works_on_dense_input(self):
for classifier in self.classifiers():
self.assertClassifierWorksWithSparsity(classifier, 'dense')
self.assertClassifierPredictsProbabilities(classifier, 'dense')
def test_if_embedding_works_with_cross_validation(self):
for classifier in self.classifiers():
self.assertClassifierWorksWithCV(classifier)
if __name__ == '__main__':
unittest.main()
