Inverse \(M\)-matrices and ultrametric matrices (Q405582)

As it is said in the Preface, this book collects the authors' work, done over 20 years, linking ultrametric matrices with potential theory of finite state Markov chains. The text establishes relations between the concepts of inverse \(M\)-matrices, ultrametric and generalized ultrametric matrices, inverse row diagonally dominant matrices and discrete potential theory. It starts with the result that ultrametric matrices are inverse \(M\)-matrices, leading to the inverse \(M\)-matrix problem, which is the characterization of nonnegative matrices whose inverses are \(M\)-matrices. This problem is then treated in Chapter 2 using probability and potential theory. Chapter 3 is about ultrametric matrices and more combinatorial in flavour, as it relates the concept to partitions and graphs (namely trees). Chapter 4 develops further this study of the relations between graphs and matrices, studying the incidence graph of inverses of ultrameric and generalized ultrametric matrices and relating these to Markov chains. Chapter 5 turns to probability theory and stochastic analysis, still keeping with the aim of describing the properties of inverse \(M\)-matrices. The concept of conditional expectation is defined and used to find a decomposition of a class of inverse \(M\)-matrices, the weakly filtered \(M\)-matrices including ultrametric and generalized ultrametric matrices. The chapter includes an algorithm that determines if a weakly filtered matrix is an inverse \(M\)-matrix. Chapter 6 deals with stability properties under Hadamard functions for the class of inverse \(M\)-matrices.