LabelEncoder

Encodes categorical labels to contiguous integer indices $[0,\, n_{\text{classes}})$, suitable for use as targets in classification models.

Mapping

$$f: \mathcal{L} \to \{0, 1, \ldots, |\mathcal{L}| - 1\}$$

where $\mathcal{L}$ is the set of unique labels sorted in ascending order. The mapping uses sorted order and binary search for $O(\log |\mathcal{L}|)$ lookup per label.

The inverse mapping inverse_transform recovers the original labels from their integer encoding.

Mirrors sklearn.preprocessing.LabelEncoder.

Constructor

Skigen::LabelEncoder<Label> encoder;

Template	Default	Description
Label	int	Label type

Methods

Method	Description
fit(y)	Learn label-to-index mapping
transform(y)	Encode labels to integers
fit_transform(y)	Fit and transform in one call
inverse_transform(y)	Decode back to original labels

Fitted Attributes

Accessor	Type	Description
classes()	std::vector<Label>	Sorted unique classes
n_classes()	Eigen::Index	Number of classes

Example

#include <Skigen/Preprocessing>

Skigen::LabelEncoder encoder;
auto encoded = encoder.fit_transform(labels);
auto decoded = encoder.inverse_transform(encoded);