Two reverse inequalities associated with Tsallis relative operator entropy via generalized Kantorovich constant and their applications
Autor: | Takayuki Furuta |
---|---|
Rok vydání: | 2006 |
Předmět: |
Numerical Analysis
Algebra and Number Theory Tsallis relative entropy Mathematical analysis Hilbert space Positive-definite matrix Linear map Combinatorics symbols.namesake Generalized Kantorovich constant Operator (computer programming) Relative operator entropy symbols Umegaki relative entropy Discrete Mathematics and Combinatorics Geometry and Topology Tsallis relative operator entropy Specht ratio Mathematics |
Zdroj: | Linear Algebra and its Applications. 412(2-3):526-537 |
ISSN: | 0024-3795 |
DOI: | 10.1016/j.laa.2005.07.013 |
Popis: | Recently Tsallis relative operator entropy Tp(A∣B) and Tsallis relative entropy Dp(A∥B) are discussed by Furuichi–Yanagi–Kuriyama. We shall show two reverse inequalities involving Tsallis relative operator entropy Tp(A∣B) via generalized Kantorovich constant K(p). As some applications of two reverse inequalities, we shall show two trace reverse inequalities involving −Tr[Tp(A∣B)] and Dp(A∥B) and also a known reverse trace inequality involving the relative operator entropy S^(A|B) by Fujii–Kamei and the Umegaki relative entropy S(A, B) is shown as a simple corollary. We show the following result: Let A and B be strictly positive operators on a Hilbert space H such that M1 I ⩾ A ⩾ m1 I > 0 and M2 I ⩾ B ⩾ m2 I > 0. Put m=m2M1, M=M2m1, h=Mm=M1M2m1m2>1 and p ∈ (0, 1]. Let Φ be normalized positive linear map on B(H). Then the following inequalities hold: (i)1-K(p)pΦ(A)♯pΦ(B)+Φ(Tp(A|B))⩾Tp(Φ(A)|Φ(B))⩾Φ(Tp(A|B)) and (ii)F(p)Φ(A)+Φ(Tp(A|B))⩾Tp(Φ(A)|Φ(B))⩾Φ(Tp(A|B)), where K(p) is the generalized Kantorovich constant defined by K(p)=(hp-h)(p-1)(h-1)(p-1)p(hp-1)(hp-h)p and K(p) ∈ (0, 1] and F(p)=mpphp-hh-11-K(p)1p-1⩾0. In addition, let A and B be strictly positive definite matrices, (iii)1-K(p)p(Tr[A])1-p(Tr[B])p+Dp(A‖B)⩾-Tr[Tp(A|B)]⩾Dp(A‖B) and (iv)F(p)Tr[A]+Dp(A‖B)⩾-Tr[Tp(A|B)]⩾Dp(A‖B). In particular, both (iii) and (iv) yield the following known result: logS(1)Tr[A]+S(A,B)⩾-Tr[S^(A|B)]⩾S(A,B), where S(1)=h1h-1elogh1h-1 is said to be the Specht ratio and S(1) > 1. |
Databáze: | OpenAIRE |
Externí odkaz: |