tsgm.dataset

Module Contents

class DatasetProperties(N: int, D: int, T: int, variables: DatasetProperties.__init__.T[DatasetProperties.__init__.T] = None)[source]

Stores the properties of a dataset. Along with dimensions it can store properties of the covariates.

Parameters:
  • N (int) – The number of samples.

  • D – The number of dimensions.

  • T – The number of timestemps.

  • variables (list) – The properties of each covariate.

class Dataset(x: tsgm.types.Tensor, y: tsgm.types.Tensor, metadata: Dict | None = None)[source]

Bases: DatasetProperties

Wrapper for time-series datasets. Additional information is stored in Distribution metadata field.

Parameters:
  • x (tsgm.types.Tensor) – The matrix of time series with dimensions NxDxT

  • y (tsgm.types.Tensor) – The lables of a time series.

  • metadata – Additional info for the dataset.

property X: tsgm.types.Tensor[source]

Returns the time series tensor in format: n_samples x seq_len x feat_dim.

property y: tsgm.types.Tensor[source]

Returns labels tensor.

property Xy: tuple[source]

Returns a tuple of a time series tensor and labels tensor.

property Xy_concat: tsgm.types.Tensor[source]

Returns a concatenated time series and labels in a tensor. Output shape is n_sample x seq_len x feat_dim + y_dim

property shape: tuple[source]

Returns the shape of the time series in the dataset.

property seq_len: int[source]

Returns the length of sequences in the dataset.

property feat_dim: int[source]

Returns the size of feature dimension in the time series.

property output_dim: int[source]

Returns the number of classes in the dataset.

__add__(other_ds: Dataset) Dataset[source]

Returns a concatenated time series and labels in a tensor. Output shape is n_sample x seq_len x feat_dim + y_dim