Implementation:Scikit learn Scikit learn MeanShift

Knowledge Sources	Scikit_learn Scikit-learn Docs
Domains	Clustering, Density-Based Clustering
Last Updated	2026-02-08 15:00 GMT

Overview

Concrete tool for performing mean shift clustering using a flat kernel provided by scikit-learn.

Description

MeanShift is a centroid-based clustering algorithm that discovers blobs in a smooth density of samples. It works by iterating over candidate centroids, updating each to be the mean of points within a given region (the bandwidth), and then filtering near-duplicates in a post-processing stage. Seeding of initial candidates is performed using a binning technique for scalability. Unlike K-Means, MeanShift does not require specifying the number of clusters; instead the number of clusters is determined by the bandwidth parameter and data distribution.

Usage

Use MeanShift when you do not know the number of clusters ahead of time and want to discover dense regions in continuous data. It is suitable for image processing, computer vision, and spatial data analysis. The bandwidth parameter controls resolution; the estimate_bandwidth utility function can help select it automatically.

Code Reference

Source Location

• Repository: scikit-learn
• File: sklearn/cluster/_mean_shift.py

Signature

class MeanShift(ClusterMixin, BaseEstimator):
    def __init__(
        self,
        *,
        bandwidth=None,
        seeds=None,
        bin_seeding=False,
        min_bin_freq=1,
        cluster_all=True,
        n_jobs=None,
        max_iter=300,
    ):

Import

from sklearn.cluster import MeanShift

I/O Contract

Inputs

Name	Type	Required	Description
bandwidth	float or None	No	Bandwidth used in the flat kernel. If None, estimated automatically using estimate_bandwidth. Default is None.
seeds	array-like or None	No	Seeds used to initialize kernels. If None, seeds are calculated using bin seeding. Default is None.
bin_seeding	bool	No	Whether to use binning to seed initial kernel locations for scalability. Default is False.
min_bin_freq	int	No	Minimum number of points in a bin to seed it as a kernel location. Default is 1.
cluster_all	bool	No	If True, all points are clustered including orphans. If False, orphans get label -1. Default is True.
n_jobs	int or None	No	Number of parallel jobs. Default is None.
max_iter	int	No	Maximum number of iterations per seed point. Default is 300.

Outputs

Name	Type	Description
cluster_centers_	ndarray of shape (n_clusters, n_features)	Coordinates of cluster centers.
labels_	ndarray of shape (n_samples,)	Cluster labels for each sample.
n_iter_	int	Maximum number of iterations performed on any seed.

Usage Examples

Basic Usage

from sklearn.cluster import MeanShift
import numpy as np

X = np.array([[1, 1], [2, 1], [1, 0],
              [4, 7], [3, 5], [3, 6]])

clustering = MeanShift(bandwidth=2).fit(X)
print(clustering.labels_)
print(clustering.cluster_centers_)

Related Pages

• Principle:Scikit_learn_Scikit_learn_Clustering