Journal
Scientific and technical journal of information technologies, mechanics and optics
UDK004.021, 004.32.2
Issue:5 (63)
New non-parametric method of choosing the number of groups in a data set is proposed. It is based on randomized stochastic approximation algorithm with input artificial perturbation. Main features to keep convergence under almost arbitrary noise are described. Proposed method could be used for on-line clustering of dynamically changed data sets. The effectiveness is demonstrated on a wide range of simulated and real nature data sets.