Structured peer-to-peer networks through distributed nonmetric multidimensional scaling
Multidimensional Scaling (MDS) is a technique for dimensionality reduction widely used in the field of data-mining. For P2P data networks, it holds the potential to provide an intuitive way to browse and explore heterogeneous data distributed over the system. Current MDS approaches are dependent on a centralized instance (controller) and can therefore not be used in decentralized environments. In this article, a distributed algorithm for multidimensional scaling is discussed that does not need any such centralized control.
Full Text: PDF