def test_subspace_tracking_momentum(self, subspace_momentum):
W, H = ornmf(self.X,
self.rank,
method="MomentumSGD",
subspace_momentum=subspace_momentum)
compare_norms(W @ H, self.X)
with pytest.raises(ValueError,
match=f"must be a float between 0 and 1"):
_ = ornmf(self.X,
self.rank,
method="MomentumSGD",
subspace_momentum=1.9)
def test_subspace_tracking_learning_rate(self, subspace_learning_rate):
W, H = ornmf(
self.X,
self.rank,
method="MomentumSGD",
subspace_learning_rate=subspace_learning_rate,
)
compare_norms(W @ H, self.X)
def test_no_method(self):
with pytest.raises(ValueError, match=f"'method' not recognised"):
_ = ornmf(self.X, self.rank, method="uniform")
def test_subspace_tracking(self):
W, H = ornmf(self.X, self.rank, method="MomentumSGD")
compare_norms(W @ H, self.X)
def test_corrupted_robust(self):
W, H = ornmf(self.Y, self.rank, method="RobustPGD")
compare_norms(W @ H, self.X)
def test_corrupted_default(self):
W, H = ornmf(self.Y, self.rank)
compare_norms(W @ H, self.X)
def test_store_error(self):
Xhat, Ehat, W, H = ornmf(self.X, self.rank, store_error=True)
compare_norms(Xhat, self.X)
assert Xhat.shape == self.X.shape
assert Ehat.shape == self.E.shape
def test_batch_size(self):
W, H = ornmf(self.X, self.rank, batch_size=2)
compare_norms(W @ H, self.X)
assert W.shape == self.U.shape
assert H.shape == self.V.T.shape
def test_default(self, project):
W, H = ornmf(self.X, self.rank, project=project)
compare_norms(W @ H, self.X)
assert W.shape == self.U.shape
assert H.shape == self.V.T.shape