Mean Shift

labels/tags

Tagged according to tag list
clustering

Principle

Mean Shift clustering belongs to the class of modal clustering methods where the arbitrarily shaped clusters are defined in terms of the basins of attraction to the local modes of the data density, created by the density gradient ascent paths. Most studies on the Mean Shift clustering have focused on kernel versions. The latter authors compared Gaussian, Cauchy and generalized Epanechnikov kernels to study the behaviour of tuning parameters of Mean Shift clustering. Mean-shift algorithm and its variants consist in two major steps as described in Algorithm 10. The first one is the density gradient ascent, it is generally the most computationally intensive. This gradient ascent can be computed in different ways. In the traditional characterization of the Mean Shift, their gradient ascent paths are computed from successive iterations of kernel applications on each data point.

Scalability

Complexity is in O(n^2). Where :

n is the number of data points

Input

A collection of continuous vectors.

Parameters

A Kernel.
Stopping criteria.
A clustering algorithm.

Ouput format

Depends on clustering technique.

Associated visualization

Depends.