Distributed Data Clustering by Neural Network Algorithms

The paper describes an analysis of the possibilities of applying Kohonen self-organizing maps for distributed data. We consider two strategies of distributed data analysis: with intermediate synchronization of results and with synchronization after full or partial construction of maps. Each strategy is considered taking into account both horizontal and vertical data distribution. We provide recommendations for the choice of a strategy depending on the type of data distribution. The obtained conclusions are confirmed by the results of an experiment.