scikit-learn-extra

0.3.0 verified Mon Apr 27 auth: no python maintenance

A set of additional estimators and tools for scikit-learn, including K-Medoids, EigenPro, and Fastfood. Current version: 0.3.0 (released Oct 2021). Releases are infrequent; the project is in maintenance mode.

pip install scikit-learn-extra

Common errors

error ModuleNotFoundError: No module named 'sklearn_extra' ↓

cause The package scikit-learn-extra is not installed.

fix

Run 'pip install scikit-learn-extra'.

error ImportError: cannot import name 'KMedoids' from 'sklearn.cluster' ↓

cause KMedoids is not part of scikit-learn core; it lives in sklearn_extra.

fix

Use 'from sklearn_extra.cluster import KMedoids'.

error ValueError: The number of clusters must be less than the number of samples. ↓

cause KMedoids requires n_clusters < n_samples.

fix

Set n_clusters to a value less than the number of data points.

Warnings

gotcha KMedoids was originally in sklearn_extra.cluster; ensure you import from sklearn_extra, not sklearn. ↓

fix Use 'from sklearn_extra.cluster import KMedoids'.

gotcha The library may not be actively maintained; check for compatibility with newer scikit-learn versions (>=1.0). ↓

fix Pin scikit-learn version to <=1.0.2 if issues arise, or consider alternatives like sklearn.cluster.KMeans.

deprecated Kernel approximation estimators (EigenPro, Fastfood) may behave differently or be removed in future versions. ↓

fix Use scikit-learn's RBFSampler as a more stable alternative.

Imports

KMedoids
wrong
```
from sklearn.cluster import KMedoids
```
correct
```
from sklearn_extra.cluster import KMedoids
```
KMedoids is not in scikit-learn core; it was removed from sklearn_extra after version 0.2.0?

EigenPro

from sklearn_extra.kernel_approximation import EigenPro

Fastfood

from sklearn_extra.kernel_approximation import Fastfood

Quickstart

Fit K-Medoids clustering on a small dataset and print cluster labels.

from sklearn_extra.cluster import KMedoids
import numpy as np
X = np.array([[1, 2], [1, 4], [1, 0], [4, 2], [4, 4], [4, 0]])
kmedoids = KMedoids(n_clusters=2, random_state=0).fit(X)
print(kmedoids.labels_)