def _test_label_model(self, data, test_acc=True, mts=False):
if mts:
label_model = LabelModel(data.m, task_graph=data.task_graph,
p=data.p, deps=data.E)
else:
label_model = LabelModel(data.m, k=data.k, p=data.p, deps=data.E)
label_model.train(data.L, n_epochs=1000, print_every=200)
# Test parameter estimation error
c_probs_est = label_model.get_conditional_probs()
err = np.mean(np.abs(data.c_probs - c_probs_est))
print(f"Parameter Estimation Error={err}")
self.assertLess(err, 0.015)
# Test label prediction accuracy
if test_acc:
Y_pred = label_model.get_label_probs(data.L).argmax(axis=1) + 1
acc = np.where(data.Y == Y_pred, 1, 0).sum() / data.n
print(f"Label Prediction Accuracy={acc}")
self.assertGreater(acc, 0.95)
def _test_label_model(self, data, test_acc=True):
label_model = LabelModel(k=data.k, verbose=False)
label_model.train(
data.L,
deps=data.E,
class_balance=data.p,
n_epochs=1000,
print_every=200,
)
# Test parameter estimation error
c_probs_est = label_model.get_conditional_probs()
err = np.mean(np.abs(data.c_probs - c_probs_est))
self.assertLess(err, 0.025)
# Test label prediction accuracy
if test_acc:
Y_pred = label_model.predict_proba(data.L).argmax(axis=1) + 1
acc = np.where(data.Y == Y_pred, 1, 0).sum() / data.n
self.assertGreater(acc, 0.95)
def _test_label_model(self, data, test_acc=True):
label_model = LabelModel(k=data.k, verbose=False)
label_model.train_model(
data.L,
deps=data.E,
class_balance=data.p,
n_epochs=1000,
log_train_every=200,
)
# Test parameter estimation error
c_probs_est = label_model.get_conditional_probs()
err = np.mean(np.abs(data.c_probs - c_probs_est))
self.assertLess(err, 0.025)
# Test label prediction accuracy
if test_acc:
score = label_model.score((data.L, data.Y), verbose=False)
self.assertGreater(score, 0.95)
# Test against baseline
mv = MajorityLabelVoter()
mv_score = mv.score((data.L, data.Y), verbose=False)
self.assertGreater(score, mv_score)