Online Scaling and Preproessing

API Reference

__init__(forget: float = 0.0, to_scale: bool | ndarray = True)

The online scaler allows for incremental updating and scaling of matrices.

Parameters:

Name	Type	Description	Default
`forget`	`float`	The forget factor. Older observations will be exponentially discounted. Defaults to 0.0.	`0.0`
`to_scale`	`bool \| ndarray`	The variables to scale. `True` implies all variables will be scaled. `False` implies no variables will be scaled. An `np.ndarray` of type `bool` or `int` implies that the columns `X[:, to_scale]` will be scaled, all other columns will not be scaled. Defaults to True.	`True`

_prepare_estimator(X: ndarray)

Add derived attributes to estimator

fit(X: ndarray) -> None

Fit the OnlineScaler() Object for the first time.

Parameters:

Name	Type	Description	Default
`X`	`ndarray`	Matrix of covariates X.	required

inverse_transform(X: ndarray) -> np.ndarray

Back-transform a scaled X matrix to the original domain.

Parameters:

Name	Type	Description	Default
`X`	`ndarray`	Scaled X matrix.	required

Returns:

Type	Description
`ndarray`	np.ndarray: Scaled back to the original scale.

partial_fit(X: ndarray) -> None

Update the OnlineScaler() for new rows of X.

Parameters:

Name	Type	Description	Default
`X`	`ndarray`	New data for X.	required

transform(X: ndarray) -> np.ndarray

Transform X to a mean-std scaled matrix.

Parameters:

Name	Type	Description	Default
`X`	`ndarray`	X matrix for covariates.	required

Returns:

Type	Description
`ndarray`	np.ndarray: Scaled X matrix.

update(X: ndarray) -> None

Wrapper for partial_fit to align API.