def runExperiment(self, classifier_name):
data, test_data, vectors = testData()
all_data = (data + test_data)*3
#print all_data
maker = ClassifierMaker(vectors)
classifier = maker.make(classifier_name)
num_folds = 3
experiment = EntailmentExperiment(all_data, classifier, num_folds)
results = [experiment.runFold(fold)
for fold in range(num_folds)]
#print results
return results
def testClassifierMakerClassifiers(self):
"Check that all classifiers that can be made are valid."
data, test_data, vectors = testData()
class_values = set(x[2] for x in data)
params = {'beta': [1.0, 2.0], 'costs': [1.0], 'k': [1]}
maker = ClassifierMaker(vectors, params)
names = maker.get_names()
for name in names:
classifier = maker.make(name)
classifier.fit(data)
results = classifier.predict(test_data)
self.assertEqual(len(results), len(test_data))
self.assertTrue(set(results) <= class_values)
def testClassifierMakerClassifiers(self):
"Check that all classifiers that can be made are valid."
data, test_data, vectors = testData()
class_values = set(x[2] for x in data)
params = {'beta':[1.0, 2.0], 'costs':[1.0]}
maker = ClassifierMaker(vectors, params)
names = maker.get_names()
for name in names:
classifier = maker.make(name)
classifier.fit(data)
results = classifier.predict(test_data)
self.assertEqual(len(results), len(test_data))
self.assertTrue(set(results) <= class_values)
def runExperiment(self, classifier_name):
data, test_data, vectors = testData()
all_data = (data + test_data)*3
#print all_data
maker = ClassifierMaker(vectors, params = {'k':[1]} )
classifier = maker.make(classifier_name)
num_folds = 3
experiment = EntailmentExperiment(all_data, classifier, num_folds)
results = [experiment.runFold(fold)
for fold in range(num_folds)]
#print results
return results
def testClassifierMakerNames(self):
data, test_data, vectors = testData()
maker = ClassifierMaker(vectors)
names = maker.get_names()
self.assertGreater(len(names), 0,
"Maker should have more than one classifier")
params['dataset'] = 'wn-noun-dependencies-original'
params['vectors'] = 'nouns-deps.mi.db'
#params['classifier'] = 'widthdiff'
params['classifier']='invCLP'
if __name__ == "__main__":
print "Testing baseline function"
datadir = params['datadir']
dataset_path = os.path.join(datadir, params['dataset'] + '.json')
random.seed(abs(hash(str(params))))
with open(dataset_path) as dataset_file:
dataset = json.load(dataset_file)
vectors_path = os.path.join(datadir, params['vectors'])
print "DB path: ", vectors_path
vectors = TermDB(vectors_path)
maker = ClassifierMaker(vectors, params)
classifier = maker.make(params['classifier'])
target = np.array([p[2] for p in dataset], dtype=int)
classifier.fit(dataset)
predictions=classifier.predict(dataset)
print "Predictions:", predictions
print "Actual:", target
def testClassifierMakerNames(self):
data, test_data, vectors = testData()
maker = ClassifierMaker(vectors)
names = maker.get_names()
self.assertGreater(len(names), 0,
"Maker should have more than one classifier")
