7月7日(周二)院学术报告

报告人: Witold  Pedrycz
        Professor and Chair,Canada Research Chair,IEEE Fellow,         IFSA Fellow,
        Department of Electrical & Computer Engineering University of Alberta, Edmonton, Canada
       & Systems Research Institute, Polish Academy of Sciences, Warsaw, Poland 
 
时间:2015年7月7日(周二),下午15:30-16:30
 
地点:后主楼1223
 
题目:  Hierarchical Structure of System Modeling with Information        Granules
 
摘要: It is evident that there are no ideal numeric models of complex systems: Essentially, all models are wrong, but some are useful (G. Box, 1976). In this study, we introduce a concept of granular models, which is inherently associated with a design of 
models whose parameters are information granules rather than numeric entities. Such information granules are formally described by sets (intervals), fuzzy sets, rough sets, probabilities etc. Subsequently, the results produced by granular models are information granules, which facilitate a better representation of obtained results as well as support an interpretation of the model itself.
 
We elaborate on a number of conceptual and design aspects of granular models. In particular, it is demonstrated that granular models developed on a basis of existing numeric models leads to their substantial augmentations and results in interesting ways 
of evaluation of their performance. The proposed assessment of the quality of the model embraces two generic criteria, namely a coverage criterion of experimental data and a specificity criterion. It is shown that a hierarchy of granular models gives rise to granular models of higher type and higher order. Furthermore it implies an emergence of so-called granular outliers along with ways of their detection. The detailed discussion is focused on fuzzy rule-based models and hierarchical fuzzy clustering (especially Fuzzy C-Means). 
 
To make the presentation self-contained, some fundamental prerequisites of Granular Computing are covered: (i) concepts of higher order and higher type information granules, (ii) construction of information granules through the involvement of the principle of justifiable granularity, and (iii) optimal allocation of information granularity.
 
邀请人:于福生

 

额外信息