splits

`splits`

Classes:

Name	Description
`Fold`	A single train/validation split.

Functions:

Name	Description
`temporal_fold`	Single temporal split.
`expanding_folds`	Generate folds for expanding-window time-series cross-validation.
`sliding_folds`	Generate folds for sliding-window time-series cross-validation.

A single train/validation split.

temporal_fold(n: int, train_ratio: float, *, gap: int = 0) -> Fold

Single temporal split.

Layout::

[===== Train =====][gap][== Val ==]

Parameters:

Name	Type	Description	Default
`n`	`int`	Total number of samples.	required
`train_ratio`	`float`	Fraction of data for training, in `(0, 1)`.	required
`gap`	`int`	Points to skip between train and validation.	`0`

Returns:

Type	Description
`Fold`

Raises:

Type	Description
`ValueError`	If ratio is invalid or any split would be empty.

expanding_folds(n: int, *, initial_train_size: int, validation_size: int, stride: int | None = None, gap: int = 0) -> list[Fold]

Generate folds for expanding-window time-series cross-validation.

The training set starts at initial_train_size and grows with each fold while the validation window slides forward.

Parameters:

Name	Type	Description	Default
`n`	`int`	Total number of samples.	required
`initial_train_size`	`int`	Minimum training samples (first fold).	required
`validation_size`	`int`	Validation samples per fold.	required
`stride`	`int or None`	Step size for the validation window. Defaults to `validation_size`.	`None`
`gap`	`int`	Points to skip between train and validation.	`0`

Returns:

Type	Description
`list[Fold]`

sliding_folds(n: int, *, train_size: int, validation_size: int, stride: int | None = None, gap: int = 0) -> list[Fold]

Generate folds for sliding-window time-series cross-validation.

A fixed-size training window slides forward together with the validation window, discarding the oldest observations at each step.

Parameters:

Name	Type	Description	Default
`n`	`int`	Total number of samples.	required
`train_size`	`int`	Fixed training samples per fold.	required
`validation_size`	`int`	Validation samples per fold.	required
`stride`	`int or None`	Step size for the sliding window. Defaults to `validation_size`.	`None`
`gap`	`int`	Points to skip between train and validation.	`0`

Returns:

Type	Description
`list[Fold]`