def test_scheduler(self):
scheduler = ReduceOnWorsening()
clf = algorithms.MEMO(max_iter=20, learning_rate=.1, scheduler=scheduler)\
.fit(self.KLtr, self.Ytr)
scheduler = ReduceOnWorsening(multiplier=.6, min_lr=1e-4)
clf = algorithms.MEMO(max_iter=20, learning_rate=.1, scheduler=scheduler)\
.fit(self.KLtr, self.Ytr)
def test_callbacks(self):
earlystop_auc = callbacks.EarlyStopping(self.KLte,
self.Yte,
patience=30,
cooldown=2,
metric='roc_auc')
earlystop_acc = callbacks.EarlyStopping(self.KLte,
self.Yte,
patience=30,
cooldown=2,
metric='accuracy')
monitor = callbacks.Monitor(
metrics=[metrics.radius, metrics.margin, metrics.frobenius])
cbks = [earlystop_auc, earlystop_acc, monitor]
clf = algorithms.MEMO(max_iter=60, learning_rate=.1, callbacks=cbks)
clf = clf.fit(self.KLtr, self.Ytr)
self.assertEqual(len(monitor.history), 3)
print(monitor.objective)
self.assertEqual(len(monitor.objective), 60)
def test_theta(self):
clf = algorithms.MEMO(theta=10, max_iter=30).fit(self.KLtr, self.Ytr)
clf = algorithms.MEMO(theta=0.0001,
max_iter=30).fit(self.KLtr, self.Ytr)
def test_MEMO(self):
self.base_evaluation(algorithms.MEMO(max_iter=10))
self.base_evaluation(algorithms.MEMO(max_iter=10, learner=SVC(C=10)))
self.base_evaluation(
algorithms.MEMO(max_iter=10, learner=algorithms.KOMD(lam=1)))