def precompute_z_norm(self, data):
"""
Stores the z-norm of every possible shapelet from data in self.z_data.
:param data: list of training examples
:type data: np.array
"""
self.z_data = dict()
for w in self.windows:
for ts_id, ts in enumerate(data):
self.z_data[ts_id, w] = z_normalize(subsequences(ts, w))
def estimate_sigma_min(self):
"""
Estimates $\sigma_{min}$ by using the maximum standard deviation of shapelets in time series
without label.
"""
if self.sigma_min is None:
sigma_min = 0
for id, labels in enumerate(self.target):
if len(labels) == 0:
ts_subs = subsequences(self.data[id], min(self.windows))
sigma_min = max(sigma_min, ts_subs.std(axis=1).max())
print("sigma_min set to {}".format(sigma_min))
self.sigma_min = sigma_min
shapelet_utils.sigma_min = self.sigma_min