def test_mass(Q, T):
m = Q.shape[0]
left = np.linalg.norm(core.z_norm(core.rolling_window(T, m), 1) -
core.z_norm(Q),
axis=1)
right = core.mass(Q, T)
npt.assert_almost_equal(left, right)
def test_stumpi_self_join():
m = 3
zone = int(np.ceil(m / 4))
seed = np.random.randint(100000)
np.random.seed(seed)
T = np.random.rand(30)
stream = stumpi(T, m, egress=False)
for i in range(34):
t = np.random.rand()
stream.update(t)
comp_P = stream.P_
comp_I = stream.I_
comp_left_P = stream.left_P_
comp_left_I = stream.left_I_
ref_mp = naive.stamp(stream.T_, m, exclusion_zone=zone)
ref_P = ref_mp[:, 0]
ref_I = ref_mp[:, 1]
ref_left_P = np.empty(ref_P.shape)
ref_left_P[:] = np.inf
ref_left_I = ref_mp[:, 2]
for i, j in enumerate(ref_left_I):
if j >= 0:
D = core.mass(stream.T_[i:i + m], stream.T_[j:j + m])
ref_left_P[i] = D[0]
naive.replace_inf(ref_P)
naive.replace_inf(ref_left_P)
naive.replace_inf(comp_P)
naive.replace_inf(comp_left_P)
npt.assert_almost_equal(ref_P, comp_P)
npt.assert_almost_equal(ref_I, comp_I)
npt.assert_almost_equal(ref_left_P, comp_left_P)
npt.assert_almost_equal(ref_left_I, comp_left_I)
np.random.seed(seed)
T = np.random.rand(30)
T = pd.Series(T)
stream = stumpi(T, m, egress=False)
for i in range(34):
t = np.random.rand()
stream.update(t)
comp_P = stream.P_
comp_I = stream.I_
comp_left_P = stream.left_P_
comp_left_I = stream.left_I_
naive.replace_inf(comp_P)
naive.replace_inf(comp_left_P)
npt.assert_almost_equal(ref_P, comp_P)
npt.assert_almost_equal(ref_I, comp_I)
npt.assert_almost_equal(ref_left_P, comp_left_P)
npt.assert_almost_equal(ref_left_I, comp_left_I)
def test_stumpi_self_join():
m = 3
zone = int(np.ceil(m / 4))
seed = np.random.randint(100000)
np.random.seed(seed)
T = np.random.rand(30)
stream = stumpi(T, m)
for i in range(34):
t = np.random.rand()
stream.update(t)
right_P = stream.P_
right_I = stream.I_
right_left_P = stream.left_P_
right_left_I = stream.left_I_
left = naive.stamp(stream.T_, m, exclusion_zone=zone)
left_P = left[:, 0]
left_I = left[:, 1]
left_left_P = np.empty(left_P.shape)
left_left_P[:] = np.inf
left_left_I = left[:, 2]
for i, j in enumerate(left_left_I):
if j >= 0:
D = core.mass(stream.T_[i:i + m], stream.T_[j:j + m])
left_left_P[i] = D[0]
naive.replace_inf(left_P)
naive.replace_inf(left_left_P)
naive.replace_inf(right_P)
naive.replace_inf(right_left_P)
npt.assert_almost_equal(left_P, right_P)
npt.assert_almost_equal(left_I, right_I)
npt.assert_almost_equal(left_left_P, right_left_P)
npt.assert_almost_equal(left_left_I, right_left_I)
np.random.seed(seed)
T = np.random.rand(30)
T = pd.Series(T)
stream = stumpi(T, m)
for i in range(34):
t = np.random.rand()
stream.update(t)
right_P = stream.P_
right_I = stream.I_
right_left_P = stream.left_P_
right_left_I = stream.left_I_
naive.replace_inf(right_P)
naive.replace_inf(right_left_P)
npt.assert_almost_equal(left_P, right_P)
npt.assert_almost_equal(left_I, right_I)
npt.assert_almost_equal(left_left_P, right_left_P)
npt.assert_almost_equal(left_left_I, right_left_I)
def test_mass(Q, T):
Q = Q.copy()
T = T.copy()
m = Q.shape[0]
ref = np.linalg.norm(core.z_norm(core.rolling_window(T, m), 1) -
core.z_norm(Q),
axis=1)
comp = core.mass(Q, T)
npt.assert_almost_equal(ref, comp)
def test_mass_inf(Q, T):
T[1] = np.inf
m = Q.shape[0]
left = np.linalg.norm(core.z_norm(core.rolling_window(T, m), 1) -
core.z_norm(Q),
axis=1)
left[np.isnan(left)] = np.inf
right = core.mass(Q, T)
npt.assert_almost_equal(left, right)
def test_mass_T_nan(Q, T):
Q = Q.copy()
T = T.copy()
T[1] = np.nan
m = Q.shape[0]
ref = np.linalg.norm(core.z_norm(core.rolling_window(T, m), 1) -
core.z_norm(Q),
axis=1)
ref[np.isnan(ref)] = np.inf
comp = core.mass(Q, T)
npt.assert_almost_equal(ref, comp)
def __init__(self, T, m, excl_zone=None):
self._T = np.asarray(T)
self._T = self._T.copy()
self._T_isfinite = np.isfinite(self._T)
self._m = m
if excl_zone is None:
self._excl_zone = int(np.ceil(self._m / 4))
self._l = self._T.shape[0] - m + 1
mp = stump(T, m)
self.P_ = mp[:, 0]
self.I_ = mp[:, 1].astype(np.int64)
self.left_P_ = np.full(self.P_.shape, np.inf)
self.left_I_ = mp[:, 2].astype(np.int64)
for i, j in enumerate(self.left_I_):
if j >= 0:
D = core.mass(self._T[i : i + self._m], self._T[j : j + self._m])
self.left_P_[i] = D[0]
self._n_appended = 0
def update(self, t):
self._T[:] = np.roll(self._T, -1)
self._T_isfinite[:] = np.roll(self._T_isfinite, -1)
if np.isfinite(t):
self._T_isfinite[-1] = True
self._T[-1] = t
else:
self._T_isfinite[-1] = False
self._T[-1] = 0
self._n_appended += 1
self.P_[:] = np.roll(self.P_, -1)
self.I_[:] = np.roll(self.I_, -1)
self.left_P_[:] = np.roll(self.left_P_, -1)
self.left_I_[:] = np.roll(self.left_I_, -1)
D = core.mass(self._T[-self._m :], self._T)
T_subseq_isfinite = np.all(
core.rolling_window(self._T_isfinite, self._m), axis=1
)
D[~T_subseq_isfinite] = np.inf
if np.any(~self._T_isfinite[-self._m :]):
D[:] = np.inf
apply_exclusion_zone(D, D.shape[0] - 1, self._excl_zone)
for j in range(D.shape[0]):
if D[j] < self.P_[j]:
self.I_[j] = D.shape[0] - 1 + self._n_appended
self.P_[j] = D[j]
I_last = np.argmin(D)
if np.isinf(D[I_last]):
self.I_[-1] = -1
self.P_[-1] = np.inf
else:
self.I_[-1] = I_last + self._n_appended
self.P_[-1] = D[I_last]
self.left_I_[-1] = I_last + self._n_appended
self.left_P_[-1] = D[I_last]
def __init__(self, T, m, excl_zone=None):
"""
Initialize the `stumpi` object
Parameters
----------
T : ndarray
The time series or sequence for which the matrix profile and matrix profile
indices will be returned
m : int
Window size
excl_zone : int
The half width for the exclusion zone relative to the current
sliding window
"""
self._T = T
self._m = m
if excl_zone is not None: # pragma: no cover
self._excl_zone = excl_zone
else:
self._excl_zone = int(np.ceil(self._m / 4))
self._illegal = np.logical_or(np.isinf(self._T), np.isnan(self._T))
mp = stumpy.stump(self._T, self._m)
self._P = mp[:, 0]
self._I = mp[:, 1]
self._left_I = mp[:, 2]
self._left_P = np.empty(self._P.shape)
self._left_P[:] = np.inf
self._T, self._M_T, self._Σ_T = core.preprocess(self._T, self._m)
# Retrieve the left matrix profile values
for i, j in enumerate(self._left_I):
if j >= 0:
D = core.mass(self._T[i:i + self._m], self._T[j:j + self._m])
self._left_P[i] = D[0]
Q = self._T[-m:]
self._QT = core.sliding_dot_product(Q, self._T)