Mean-shift Clustering with Tensorflow

Writen and maintained by Phong Le ([email protected])

Description

Clustering using mean shift with Gaussian kernel.

Requirements

• python 3.5
• tensorflow 0.12

Run

To set parameters, open src/meanshift.py

n_samples = 300 # sample size
n_centroids = n_samples # number of centroids
n_updates = -1 # number of updates, -1 means run until converge (diff < 1e-5)
window_radius = .1 # window radius for the Gaussian kernels
n_gaussians = 3 # 'true' number of clusters
   

Execute

python3 src/meanshift.py