On the association between a random parameter and an observable

摘要: The association between an observable and a random parameter characterizes their joint distribution given the marginal distributions. It has been shown to be incorporated in (log-)odds ratio function. is inherent each of conditional distributions hence determines learning process formalized Bayes' theorem. paper focuses on two applications. Commonly used measures dependence, especially Kullback-Leibler distances densities interest are identified interpreted as expected values log-odds Frequently Bayesian inference based maximization utility. If utility probability density defined by logarithmic score function, can often decomposed approximately into term ``fit'' ``model complexity''. parameterization reveals that complexity'' again value i.e., measure dependence parameter. ideas illustrated throughout with examples from class conjugate exponential families.