def test_load(self):
examples = train_example.load()
print("X length: ", len(examples[0].x))
print("Y length: ", len(examples[0].y))
self.assertEqual(len(examples), 1)
def train():
examples = train_example.load()
# Split examples in train and validation using 5-fold
print("examples size:", len(examples))
wLength = len(examples[0].x[0].coref.feat)
w = np.array([0]*wLength)
w = perceptron.structured(w, examples, test=None, epochs=38, argmax=quotation.argmax, phi=quotation.phi)
return w
def calibration():
examples = train_example.load()
# Split examples in train and validation using 5-fold
print("examples size:", len(examples))
kf = KFold(len(examples), n_folds=5)
wLength = len(examples[0].x[0].coref.feat)
for trainIndex, testIndex in kf:
train = [ examples[e] for e in trainIndex ]
print("\n\ntrain size:", len(train))
validation = [ examples[e] for e in testIndex ]
print("validation size:", len(validation))
w = np.array([0]*wLength)
w = perceptron.structured(w, train, test=validation, epochs=65, argmax=quotation.argmax, phi=quotation.phi)