def setUp(self):
self.n_signals = 4
self.n_samples = 10
self.arma_orders = [(3, 2), (1, 5)]
self.dataset = RandomArmaDataset(self.n_signals, self.n_samples,
self.arma_orders)
def test_normalize_returns_same_signals_with_unit_variance(self):
dataset_norm = RandomArmaDataset(**self.kwargs, normalize=True)
for sig, sig_norm in zip(self.dataset, dataset_norm):
# noinspection PyTypeChecker
self.assertAlmostEqual(np.std(sig_norm.y), 1)
np.testing.assert_allclose(sig_norm.y, sig.y / np.std(sig.y))
np.testing.assert_allclose(sig_norm.u, sig.u)
np.testing.assert_equal(sig_norm.usage_seq, sig.usage_seq)
def setUp(self):
self.n_signals = 3
self.n_samples = 100
self.arma_orders = [(3, 2), (2, 4), (1, 5), (4, 0)]
self.kwargs = dict(n_signals=self.n_signals,
n_samples=self.n_samples,
arma_orders=self.arma_orders)
self.dataset = RandomArmaDataset(**self.kwargs)
def test_arma_kws_forwarded_to_make_random_arma(self):
dataset = RandomArmaDataset(3,
20,
arma_orders=[(1, 1), (2, 2)],
arma_kws={"bias_range": (0, 1)})
for crt_armas in dataset.armas:
for crt_arma in crt_armas:
self.assertGreater(crt_arma.bias, 1e-3)
def setUp(self):
self.n_signals = 4
self.n_samples = 10
self.arma_orders = [(3, 2), (1, 5)]
self.arma_hsmm_kws = {"min_dwell_time": 5}
self.dataset = RandomArmaDataset(self.n_signals, self.n_samples,
self.arma_orders,
**self.arma_hsmm_kws)
def test_source_is_copied_before_using(self):
dataset = RandomArmaDataset(1,
23, [(1, 2), (3, 0)],
source=sources.GaussianNoise())
sig1 = dataset[0]
sig2 = dataset[0]
np.testing.assert_equal(sig1.y, sig2.y)
np.testing.assert_equal(sig1.u, sig2.u)
np.testing.assert_equal(sig1.usage_seq, sig2.usage_seq)
def create_dataset(self, fix_scale: Optional[float]):
with mock.patch(
"bioslds.dataset.sources.fix_source_scale") as MockFixScale:
MockFixScale.side_effect = self.scales
dataset = RandomArmaDataset(
self.n_signals,
self.n_samples,
arma_orders=[(0, 0),
(0, 0)], # ARMAs simply return their inputs
fix_scale=fix_scale,
)
return dataset
def test_init_kwargs_forwarded_to_arma_hsmm(self):
start_prob = [1, 0, 0]
trans_mat = [[0, 1, 0], [0, 0, 1], [1, 0, 0]]
n_samples = 25
dataset = RandomArmaDataset(
2,
n_samples,
arma_orders=[(3, 0), (2, 1), (1, 3)],
start_prob=start_prob,
trans_mat=trans_mat,
)
for crt_sig in dataset:
np.testing.assert_equal(crt_sig.usage_seq,
np.arange(n_samples) % 3)
def setUp(self):
self.n_signals = 3
self.n_samples = 100
self.alphas = [0.99, 0.95]
self.armas = [Arma([_], []) for _ in self.alphas]
self.y0 = 1.0
self.dataset = RandomArmaDataset(
self.n_signals,
self.n_samples,
armas=self.armas,
source=sources.Constant(0),
initial_conditions=([self.y0], []),
)
def setUp(self):
self.n_signals = 3
self.n_samples = 100
self.arma_orders = [(3, 3), (2, 3)]
self.fix_scale = 1.5
self.scales = [2.3, 0.2, 1.2, 1.5, 0.2, 1.0]
with mock.patch(
"bioslds.dataset.sources.fix_source_scale") as MockFixScale:
MockFixScale.side_effect = self.scales
self.dataset = RandomArmaDataset(
self.n_signals,
self.n_samples,
arma_orders=self.arma_orders,
fix_scale=self.fix_scale,
)
self.mock_fix_scale = MockFixScale
def setUp(self):
self.n_signals = 3
self.n_samples = 32
self.dataset = RandomArmaDataset(self.n_signals, self.n_samples,
[(3, 0), (1, 1)])
def create_dataset(self, rng) -> RandomArmaDataset:
dataset = RandomArmaDataset(self.n_signals,
self.n_samples,
self.arma_orders,
rng=rng)
return dataset
def test_raises_if_no_armas_and_no_arma_orders(self):
with self.assertRaises(ValueError):
RandomArmaDataset(3, 3)
def setUp(self):
self.n_signals = 5
self.n_samples = 25
self.dataset = RandomArmaDataset(self.n_signals, self.n_samples,
[(1, 1), (2, 1)])