Functioning by analogy (Q1107480)

The paper covers some selected aspects of functioning by analogy. It is pointed out how a class of functions enabling this class of reasoning results in the framework of the theory of fuzzy measure and fuzzy integral. Following an observation that the higher values of a similarity degree \(\chi\) (x,y) of two elements x and y imply higher values of grades of belongingness (membership) of x and y to a particular class \(\omega\) (\(\mu\) (x) and \(\mu\) (y)), the resulting membership function is derived with the aid of fuzzy integrals. Considering a fuzzy reflexive concordance relation R between elements from a universe X and converting the above observation into a set of inequalities we obtain, \[ -\mu (x)[1- R(x,y)]\leq \mu (y)-\mu (x)\leq \mu (y)[1-R(x,y)] \] or \[ \min (\mu (x),R(x,y))-\mu (x)\leq \mu (y)-\mu (x)\leq \mu (y)-\min (\mu (y),R(x,y)). \] These constraints are read as follows, \[ (1)\quad \mu (y)=\sup_{x\in X}[\mu (x)R(x,y)],\quad (2)\quad \mu (y)=\sup_{x\in X}[\min (\mu (x),R(x,y))], \] respectively. Making use of fuzzy integral, (1) and (2) are equivalent to the expression \(\mu =\mu \circ g\). Then the membership function \(\mu\) results as an eigen membership function of R. A solution of the above equation with the fuzzy set \(\mu\) is discussed. A suitable choice of the fuzzy relation R is studied. Some strictly applicational aspects are also dealt with.