This chapter is devoted to a brief exposition of relative entropy, and to proofs of Csiszar's projection theorem and those on exponential families needed in Chapter 5. 10.1. Relative Entropy Let {n, B} be a measurable space, MI (n) : the space of probability measures on n, and the relative entropyl H(QIP) of Q with respect to P be defined by (10.1) H(QIP)=fClOgdQ)dQ, ifQ«P, (=00, otherwise). dP where P, Q E MI(n). Denoting the Radon-Nikodym derivative of Q with respect to P as q = qp = if, , the relative entropy H(Q I P) defined by (10.1) can be written in terms of the density function q as 1 It has appeared under various names, cf. Kullback (1959), Csiszar (1975,84) M. Nagasawa, Schrödinger Equations and Diffusion Theory
CITATION STYLE
Nagasawa, M. (1993). Relative Entropy and Csiszar’s Projection. In Schrödinger Equations and Diffusion Theory (pp. 239–252). Birkhäuser Basel. https://doi.org/10.1007/978-3-0348-8568-3_10
Mendeley helps you to discover research relevant for your work.