Glossary

Wildboar embraces the glossary of terms by scikit-learn with some additions.

time-series

The most common input to Wildboar estimators. An array-like that for which numpy.asarray will produce an array of appropriate shape, with rank 1, 2 or 3.

single time-series

A 1d-array with shape (n_timestep, ) or 2d-array with a single row or column.

univariate time-series

A 2d-array with shape (n_samples, n_timestep).

multivariate time-series

A 3d-array with shape (n_samples, n_dims, n_timestep).

variable-length time-series

A time-series were each sample or dimension can be of different length. The maximum length is given by arr.shape[-1], but each sample can have a length shorter than that.

missing-values

A missing value represented by numpy.nan.

end-of-series value

A missing value that also indicates that the time-series is variable-length, represented by wildboar.utils.variable_len.EoS. Any value with an index larger than the first EoS is assumed not to be part of the series. numpy.isnan returns True for EoS. To check for exactly EoS, use wildboar.utils.variable_len.is_end_of_series.

timestep (or n_timestep)

The length of the time series given by arr.shape[-1].

dimensions (or n_dims)

The number of dimensions of a (multivariate) time series. For 2d-arrays the number of dimensions is 1 and for 3d-arrays the number of dimensions is given by arr.shape[1].