报告一:General Bayesian learning: This talk will discuss problems with Bayesian inference as it currently stands and introduce a new formulation for updating belief distributions using the notion of loss functions. Indeed, loss functions will be seen as a fundamental replacement of the use of statistical models based on probability distributions.
时 间:2012年6月20日上午9:30-11:00
地 点:二教南五楼524会议室
报告人:Prof. Stephen Walker
报告二:Bayesian nonparametrics and conditional distributions: This talk will develop an interpretable construction of the regression model in a flexible Bayesian nonparametric setting. The key is to think of such models as being mixtures of linear regression models, for example, but where the range of applicability of the independent variable needs to be modeled explicitly.
时 间:2012年6月21日下午1:30-3:00
地 点:二教南五楼524会议室
报告人:Prof. Stephen Walker
报告三:Bayesian nonparametric estimation of a copula: Copulas have recently become popular as a modeling tool for accounting for the dependence structure of multivariate data. This talk will present here a Bayesian nonparametric methodology to estimate arbitrary bivariate copula density and prove that any bivariate copula density can be approximated by an infinite mixture of the Gaussian copula densities, the dependence structures of the pairs with standard normal marginals. The proposed approach extends the range of estimation since there is no need to any assumption of the copula family models.
时 间:2012年6月21日下午3:00-4:30
地 点:二教南五楼524会议室
报告人:Prof. Xue Wang
CV: Stephen G. Walker obtained his first degree in Mathematics from the University of Oxford, and obtained his PhD from Imperial College, London, in 1995. He was promoted to Professor of Statistics at Bath in 2000 and is currently at the University of Kent. He is AE