Subadditivity of Entropy for Stochastic Matrices

From MaRDI portal





The entropy of an irreducible row-stochastic matrix \(P=(p_{ij})\) with the associated equilibrium probability distribution \(p=(p_i)\) is \(H(P)=-\sum_{i,j} p_i p_{ij} \ln (p_{ij})\), the entropy of the associated Markov shift. For matrices \(A,B,C\) with common stationary vector it is proved \[ H(C)\leq H(BC) \leq H(B)+H(C), \qquad H(ABC)+H(B)\leq H(AB)+H(BC). \]











This page was built for publication: Subadditivity of Entropy for Stochastic Matrices

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3151297)