Ultrametric Matrices and Induced Markov Chains
Source: Advances in Applied Mathematics, Volume 17, Number 2, June 1996 , pp. 169-183(15)
Publisher: Academic Press
Abstract:We study non-singular ultrametric matrices A . These kinds of matrices are restrictions of non-singular tree matrices. The structure of A -1 allows us to associate to A some substochastic kernels P . We are able to describe the graph of P , in particular those vertices which lose mass. Our main tools are an algebraic decomposition lemma of A -1 and the study of the minimal tree matrix associated to A .
Document Type: Research Article
Publication date: June 1, 1996