SOM variants for topological horizontal collaboration

In this paper, we focus on collaborative clustering methods based on topological approaches, such as self-organizing maps (SOM) and self-organizing maps based on a locally adapting neighborhood radii (AdSOM). So far, the methods of clustering carried out on a single dataset, but current applications require datasets distributed among multiple sites. Thus, the communication between various datasets is required. The basic concept of collaborative clustering is to collaborate by exchanging information on their consensus. The strength of collaboration, or confidence, is specified by a coefficient of collaboration. Collaboration can be made vertical, horizontal and hybrid: The horizontal collaboration is applied for datasets that represent the same individuals but with different feature spaces. However, the vertical collaboration occurs when the datasets contain different individuals and same variables. Hybrid collaboration allows combining the horizontal and the vertical approaches. In this present work, we are particularly interested in horizontal topological collaborative clustering.