On the max-generalized mean powers of a fuzzy matrix (Q2270252)

The paper deals with asymptotic properties of max-mean powers of fuzzy matrices. Since the max-mean product is not associative, then a power sequence is not unique (here the right hand generated max-mean powers are considered). The max-arithmetic mean composition of fuzzy relations was introduced by \textit{J.~C.~Bezdek} and \textit{J. D. Harris} [Fuzzy Sets Syst. 1, 111--127 (1978; Zbl 0442.68093)]. Recently, the authors applied this composition for examination of powers of fuzzy matrices [Fuzzy Sets Syst. 158, No.~22, 2516--2522 (2007; Zbl 1155.15020)]. Now, they generalize their own results for the case of the max-mean composition, where the weighted power mean is used. In particular, the considered power sequence is always convergent, the limit matrix has constant columns (Theorem 1), and the initial matrix has the unique max-mean row eigenvector (Theorem 7). These results have an application in consideration of max-mean fuzzy Markov chains, what generalizes results of \textit{K. E. Avrachenkov} and \textit{E.~Sanchez} [Fuzzy Optim. Decis. Mak. 1, No.~2, 143--159 (2002; Zbl 1091.91502)] (where the case of the arithmetic mean was considered). The last result (Theorem 9) states a stability of matrix powers with respect to small perturbations of the initial matrix (the transition matrix in fuzzy Markov process).