Skip to content
This repository has been archived by the owner on Nov 14, 2022. It is now read-only.
/ quickshift Public archive

A clustering algorithm that first finds the high-density regions (cluster-cores) of the data and then clusters the remaining points by hill-climbing. Such seedings act as more stable and expressive cluster-cores than the singleton modes found by popular algorithm such as mean shift. (https://arxiv.org/abs/1805.07909)

License

Notifications You must be signed in to change notification settings

google/quickshift

Quickshift++

This is not an officially supported Google product

Density-based clustering algorithm based on mode-seeking.

Usage

Initializiation:

QuickshiftPP(k, beta)

k: number of neighbors in k-NN

beta: fluctuation parameter which ranges between 0 and 1.

Finding Clusters:

fit(X)

X is the data matrix, where each row is a datapoint in euclidean space.

fit performs the clustering. The final result can be found in QuickshiftPP.memberships.

Example (mixture of two gaussians):

from QuickshiftPP import *
import numpy as np

X = [np.random.normal(0, 1, 2) for i in range(100)] + [np.random.normal(5, 1, 2) for i in range(100)]
y = [0] * 100 + [1] * 100

# Declare a Quickshift++ model with tuning hyperparameters.
model = QuickshiftPP(k=20, beta=.5)

# Compute the clustering.
model.fit(X)
y_hat = model.memberships

from sklearn.metrics.cluster import adjusted_rand_score, adjusted_mutual_info_score
print("Adj. Rand Index Score: %f." % adjusted_rand_score(y_hat, y))
print("Adj. Mutual Info Score: %f." % adjusted_mutual_info_score(y_hat, y))

Install

This package uses distutils, which is the default way of installing python modules.

To install for all users on Unix/Linux:

sudo python setup.py build; python setup.py install

Dependencies

python 2.7, scikit-learn

About

A clustering algorithm that first finds the high-density regions (cluster-cores) of the data and then clusters the remaining points by hill-climbing. Such seedings act as more stable and expressive cluster-cores than the singleton modes found by popular algorithm such as mean shift. (https://arxiv.org/abs/1805.07909)

Resources

License

Code of conduct

Security policy

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published