Authors: F. Cazals and G. Carriere

Introduction

This package implements seeding methods, i.e. intialization methods for clustering algorithms, such as k-means. These methods provide initial positions for cluster centers at a fixed runtime cost. Additionaly, these methods can also be used to provide initial positions for components in mixture models.

Implemented seed selection methods include:

• Random seed selection
• Minimax seed selection
• Greedy K-means++ seed selection

We also include reselection seeding methods, designed to improve the initialized seeds at a fixed cost, to create multi-step initialization procedures. ADD REFERENCE TO PAPER

Implemented seed reselection methods include:

• Greedy K-means++ seed reselection using cluster seeds SSE
• Greedy K-means++ seed reselection using cluster centers of mass SSE