講座內(nèi)容：In system modeling, we often develophierarchies of models to cope with the complexity of the underlying system orprocess to be modeled and realize different views at the system positioned atvarious levels of generality/abstraction to address specific needs of users.Granular Computing arises here as a viable alternative: information granules builtat a suitable level of generality and thus help establish a suitable modelingperspective and are regarded as crucial building blocks of ensuing granularmodels.

Clustering (fuzzy clustering, roughclustering, etc.) offers a comprehensive framework for constructing informationgranules as data-driven constructs. In their design, there is a need for aparadigm shift: information granules have to be developed in the presence ofsources of knowledge of different nature (quite often auxiliary to the knowledgeresiding within locally available data). This gives rise to the concept of knowledge-oriented and content-richinformation granules.

A fundamental question emerges aboutan efficient conceptual setup in which the sources of auxiliary knowledge,especially functionality requirements, may contribute to the designed systems.Both the content and the underlying semantics of obtained information granulesrequire a through exploration so that their properties could be effectivelyutilized in the realization of the model.

The objective of this presentation isto discuss fundamental ways of augmenting the existing conceptual andalgorithmic setup of building information granules (clusters) by accommodatingdomain knowledge and forming mechanisms of usage of the content of informationgranules in system modeling. Specific schemes of condition-based clustering,collaborative clustering and higher-order clustering are elaborated on.Furthermore we discuss emergence of granular models involving informationgranules of higher type.