A heuristic for the automatic parameterization of the spectral clustering algorithm

Pierrick Bruneau (1)

1 CRP-GL - Centre de Recherche Public Gabriel Lippmann, Département Informatique, Systèmes et Collaboration

Finding the optimal number of groups in the context of a clustering algorithm is a known as a difficult problem. In this article, we describe and evaluate a heuristic thereof for the spectral clustering algorithm. Our method is deterministic, and remarkable by its low computational burden. We show its effectiveness in most cases. Some limits are identified though, and serve to the formulation of perspectives to this work.

Mots clés: spectral clustering; automatic parameterization

Domaines: Apprentissage [cs.LG]

Pierrick Bruneau. A heuristic for the automatic parameterization of the spectral clustering algorithm. 2013. ⟨hal-00868416⟩