Clustering and Classification Using Recursive Mixture Estimation

The proposed project deals with the issues of clustering and classification from the viewpoint of Bayesian methodology and using the recursive mixture estimation theory. The project is directed at systematic development of this theory with the main potential in (i) further extensions of the pointer model (static data-dependent, dynamic data-dependent, etc.), indicating the active component, in combination with (ii) components determined for various types of data (static, dynamic, different distributions, mixtures of different distributions, etc.) and (iii) development of new recursive algorithms which will allow a real-time non-iterative data mining. The ideas to be realized during the proposed project will contribute to development of a novel real-time systematic tool which covers the tasks of clustering and classification (completely from theory to software).