The Study And Application Of Hybrid Fuzzy Semantic Cell

The expression of conceptual entities often comes with some ambiguity,which is implied in semantic expression.The use of appropriate conceptual model to express fuzzy semantics is of grate importance.Fuzzy semantic cell,as the representation unit of the smallest fuzzy concept,plays an important role in data mining,machine learning and knowledge discovery.In the conceptual space(discourse)?,fuzzy semantic cell L =<P,d,?>is called 'about pi','similar to pi','close to pi' semantic label,where P,d,? are the prototype of label,a distance function on ?,and a probability density function on[0,+?).In fuzzy semantic cell learning,we need to pay attention to three factors:semantic coverage,clarity of description,and ambiguity of description.Therefore,the ambiguous semantic cell learning goal is naturally related to the maximum coverage,the most typical and the maximum fuzzy entropy of the three indicators above.In this paper,mixed fuzzy semantic cell is built based on fuzzy semantic cell learning.The goal of fuzzy semantic cell learning is to find the best L to describe the data set of a certain concept while the mixed fuzzy semantic cell is further expanded on this basis.Consider a collection of several related conceptsLA = {L1,L2,...,Ln},each of which corresponds to fuzzy semantic cell Li which describe the data set of concept.Mixed fuzzy cell learning is to find a set of the most appropriate weight parameters to describe the impact or the degree of importance of a concept in the set.Referring to the principle of the learning of fuzzy semantic cell,we need to redefine and compute the two numeric characteristics of semantic cell:expectation granularity R and fuzzy entropy H.Finally,the learning of mixed fuzzy semantic cell is transformed into a nonlinear constrained optimization problem.