The semiring of matrices over a finite chain (Q942323)

Given a real \(n\times n\) matrix with \(m\) different entries, we can order these elements as \(a_0<a_1<\cdots<a_{m-1}\) and obtain a matrix with entries from the chain \(L_m=\{0,1,\dots,m-1\}\) if each element is replaced by its index. The result is a matrix in the semiring \(M(L_m)\). A cut matrix is a (real) matrix where all elements below some threshold are replaced by 0 and the others by 1. This idea can be generalized. Take a matrix in \(M(L_m)\) and replace all entries below some lower bound by zero, all entries above some upper bound by \(k-1\) and keeping the in-between elements. This defines a map from \(M(L_m)\) to \(M(L_k)\). This mapping is an order preserving semiring homomorphism that respects the addition and multiplication in the chains and hence also the matrix operations. An associated lifting operation is defined from \(M(L_k)\) to \(M(L_m)\), again an order preserving semiring homomorphism. The composition is a mapping from \(M(L_m)\) to \(M(L_m)\) which will give a ``truncated'' version of the original matrix. By fixing \(k\) but translating the upper and lower cut-off bounds, a kind of sliding window over \(L_m\) is defined. Each window gives a ``truncated'' version of \(A\in M(L_m)\). This leads to a decomposition of \(A\) writing it as a sum of truncated matrices summed over the sliding windows. If the original real matrix is a (real) fuzzy matrix with \(m\) distinct elements (all in \([0,1]\)), then by mapping it to \(A\in M(L_m)\), it is shown that its index and period equals the index and period of \(A\). Hence the results presented can help in analyzing fuzzy matrices.