Clustering of data : a physicist’s (novel) approach.
Eytan Domany (Weizmann Institute of Science - Israël)

Jeudi 10 octobre 1996

Clustering is an important technique in exploratory data analysis, with wide-ranging applications. The aim is to partition data according to natural classes present in it, assigning data points that are "more similar" to the same "cluster". We solved this ill-posed problem without making any assumptions about the structure of the data, by using a physical system as an analog computer.

The physical system we use is a disordered (granular) magnet. The method was tested successfully on a variety of artificial and real-life problems, such as classification of flowers, processing of satellite images, speech recognition and identification of textures.