nami_toys.standardise#

Classes

Standardiser(mean, std)

Shift-and-scale transform fitted from data.
class nami_toys.standardise.Standardiser(mean, std)[source]#

Bases: object

Shift-and-scale transform fitted from data.

Parameters:

  • mean (Tensor) – Per-feature mean, shape (d,).

  • std (Tensor) – Per-feature standard deviation, shape (d,).

classmethod fit(x, *, eps=1e-08)[source]#

Compute mean and std from a data tensor (N, d).

Features with zero variance are given std = 1 to avoid division by zero.

Return type:

Standardiser

Parameters:
inverse(x)[source]#

Map standardised x back to the original scale.

Return type:

Tensor

Parameters:

x (Tensor)

mean: Tensor#
std: Tensor#
transform(x)[source]#

Standardise x to zero mean and unit variance.

Return type:

Tensor

Parameters:

x (Tensor)

transform_dataset(ds)[source]#

Return a copy of ds with standardised features.

Return type:

ToyDataset

Parameters:

ds (ToyDataset)