RFC: add on-line k-means clustering

[kgryte]

Checklist

Please ensure the following tasks are completed before submitting a feature request.

• Read and understood the Code of Conduct.
• Searched for existing issues and pull requests.
• The issue name begins with RFC:.

Description

Description of the feature request.

This RFC proposes to add an implementation for on-line k-means clustering with automatic determination of k. This stands in contrast to traditional k-means clustering implementations which require the number of clusters k to be known a priori.

Related Issues

Does this feature request have any related issues?

No.

Questions

Any questions for reviewers?

• What should the package name be and where should the package reside in the project? The implementation performs incremental updates and, thus, fits within the realm of ml/incr; however, ml/incr/kmeans is currently allotted to an implementation requiring a fixed number of k clusters. This implementation is adaptive. So possibly the name should be ml/incr/adaptive-kmeans or ml/incr/akmeans?

Other

Any other information relevant to this feature request? This may include screenshots, references, sample output, and/or implementation notes.

Reference literature:

• An algorithm for online k-means clustering
• presentation