Variable threshold concept lattices (Q2456458): Difference between revisions

A fuzzy formal context may be defined as a triple \((U,A,\tilde I)\) with \(U\) and \(A\) finite sets (of objects and of attributes, respectively), and \(\tilde I\colon U \times A \to L\) a binary fuzzy relation, where \(L\) is the unit interval \([0,1]\). Both the crisp powerset \( P(Z)\) and the set \(L^Z\) of fuzzy subsets of any set \(Z\) are considered as posets ordered by inclusion. Now, if \(M \in \{P(U),U^L\}\), \(N \in \{P(A),A^L\}\), \(f\colon M \to N\), \(g\colon N \to M\), then a pair \((X,B) \in M \times N\) is called a fuzzy concept if \(f(X) = B\) and \(g(B) = X\). \textit{S. Elloumi, J. Jaam, A. Hasnah, A. Jaoua} and \textit{I. Nafkha} suggested in [Inf. Sci. 163, No.~4, 253--262 (2004; Zbl 1076.68085)] to employ a treshold \(\delta \in (0,1]\) to obtain what is called a variable threshold concept by adjusting the parameter \(\delta\). The derivation operators \(f\) and \(g\) used in that paper do not form a Galois connection, hence, several important properties of classical concept lattices do not hold for such \(\delta\)-fuzzy concept lattices. In the paper under review, three other pairs of derivation operators are introduced between \(P(U)\) and \(P(A)\), \(P(U)\) and \(L^A\), \(L^U\) and \(L^A\). All of them are Galois connections; moreover, the generated variable threshold concept lattices are considerably smaller in size than the traditional fuzzy concept lattice. Properties and relationships between these types of lattices are examined, and an illustrative example using a formal concept from [loc. cit.] is given. Reviewer's remark: The authors' constructions are analysed by \textit{R. Belohlavek} in the paper [Inf. Sci. 177, No.~15, 3186--3191 (2007; Zbl 1119.06004)], where the main properties of the new derivation operators are obtained as consequences of recent results obtained by him and S. Pollandt.