skx 0.0.21

scikit-learn compatible extensions.

skx

scikit-learn compatible extensions for supervised learning on tabular data.

Documentation

See docs.sk-x.org.

Requirements

• Python >= 3.12
• scikit-learn ~= 1.7.0
• numpy >= 2.0, scipy >= 1.13, pandas >= 2.0

Install

pip install skx

Usage

Augmentation functions

import numpy as np
from skx.augmentation import gaussian_augment

X = np.array([[1, 2], [3, 4], [5, 6]])
y = np.array([1, 2, 3])
X_aug, y_aug, sample_weight = gaussian_augment(X, y, factor=2.0, y_std=0.1)

Augmentation meta-estimators

from sklearn.linear_model import LinearRegression
from skx.augmentation import GaussianAugmentedRegressor

reg = GaussianAugmentedRegressor(LinearRegression(), factor=10.0, y_std=0.1)
reg.fit(X, y)
pred = reg.predict([[2, 3]])

Ensembles

from sklearn.linear_model import LinearRegression
from skx.ensemble import MixtureOfExpertsRegressor

moe = MixtureOfExpertsRegressor(
    estimator=LinearRegression(), n_estimators=5, split="kmeans", n_clusters=3
)
moe.fit(X, y)
pred = moe.predict(X)

Neural network

from skx.neural_network import LinearScalingMLPRegressor

mlp = LinearScalingMLPRegressor(
    n_hidden_layers=2,
    hidden_layer_width=50,
    shrink_factor=0.5,
    max_iter=100,
)
mlp.fit(X, np.column_stack([y, y]))  # multi-output example