gluonts.nursery.spliced_binned_pareto.data_functions module

gluonts.nursery.spliced_binned_pareto.data_functions.add_spikes(ts: torch.Tensor, only_upper_spikes: bool = False)[source]

Adds spikes to 15% of the time series in the form of heavy-tailed (Generalized Pareto) realizations

Parameters
  • ts – time series

  • only_upper_spikes – boolean to indicate upper-tailed or two-tailed spikes

gluonts.nursery.spliced_binned_pareto.data_functions.add_spikes_asymmetric(ts: torch.Tensor, xi: List[float] = [0.02, 0.04])[source]

Adds spikes to 15% of the time series in the form of heavy-tailed (Generalized Pareto) realizations

Parameters
  • ts – time series

  • xi – [float, float], GenPareto heaviness parameter for [lower, upper] noise respectively

gluonts.nursery.spliced_binned_pareto.data_functions.create_ds(num_points: int, t_dof: int = 10, noise_mult: float = 0.25, points_per_sinusoid: int = 100, magnitude_sin: float = 1)[source]

Creates noisy sinusoid. (Noise distributed as student t with degrees of freedom = t_dof.) Returns tensor of shape (1, 1, num_points).

Parameters
  • num_points – int, number of points in the dataset.

  • t_dof – int, degrees of freedom for student t distribution.

  • noise_mult – float, standard deviation.

  • points_per_sinusoid – int, datapoints per sine period

  • magnitude_sin – float, magnitude of sine amplitude

gluonts.nursery.spliced_binned_pareto.data_functions.create_ds_asymmetric(num_points: int, t_dof: List[float] = [10, 10], noise_mult: List[float] = [0.25, 0.25], xi: List[float] = [0.02, 0.04], points_per_sinusoid: int = 100, magnitude_sin: float = 1)[source]

Creates noisy sinusoid. (Noise distributed as student t with degrees of freedom = t_dof.) Returns tensor of shape (1, 1, num_points).

Parameters
  • num_points – int, number of points in the dataset.

  • t_dof – [int, int], degrees of freedom for Students’-t distribution for [lower, upper] noise respectively.

  • noise_mult – [float, float], standard deviation for [lower, upper] noise respectively.

  • points_per_sinusoid – int, datapoints per sine period

  • magnitude_sin – float, magnitude of sine amplitude

gluonts.nursery.spliced_binned_pareto.data_functions.create_ds_iid(num_points: int, noise_mult: float = 0.25)[source]

Creates heavy-tailed gaussian iid. Returns tensor of shape (1, 1, num_points).

Parameters
  • num_points – int, number of points in the dataset.

  • noise_mult – float, standard deviation