Abstract
The aim of clustering is to partition data according to natural classes present in it. We proposed recently a method that makes no explicit assumption about the structure of the data and under very general and natural assumptions solves the clustering problem by evaluating thermal properties of a disordered (granular) magnet. The method was tested successfully on a variety of artificial and real-life problems; here we emphasize its application to analyze results obtained by a novel method of computer vision. The combination of these two techniques provides a powerful tool that succeeded to cluster properly 90 images of 6 objects on the basis of their pairwise dissimilarities. These dissimilarities, which constitute a highly non-metric set of pairwise distances between the images, form the input for clustering. A hierarchical organization of the images that agrees with human intuition, was obtained without assigning to the images coordinates in some abstract space.
| Original language | English |
|---|---|
| Pages (from-to) | 5-12 |
| Number of pages | 8 |
| Journal | Computer Physics Communications |
| Volume | 121 |
| DOIs | |
| State | Published - Sep 1999 |
| Event | Proceedings of the 1998 Europhysics Conference on Computational Physics (CCP 1998) - Granada, Spain Duration: 2 Sep 1998 → 5 Sep 1998 |
Bibliographical note
Funding Information:We thank Shai Wiseman, Gaddy Getz and Noam Shental for most pleasant collaborations. The fMRI work required considerable patience on the part of Kalanit Grill-Spector and Rafi Malach. Discussions with several colleagues are warmly acknowledged; Ido Kanter, Michele Vendruscolo and David Mukamel in particular. This research was supported in part by the GIF – Germany Israel Science Foundation.