The horseshoe prior for edge-preserving Bayesian inversion
数学学科创建110周年系列报告
报告题目(Title):The horseshoe prior for edge-preserving Bayesian inversion
报告人(Speaker):董艺秋 副教授(丹麦科技大学)
地点(Place):后主楼1220
时间(Time):2025年10月24日(周五)15:00-16:00
邀请人(Inviter):刘君
报告摘要
In many large-scale inverse problems characterization of sharp edges in the solution is desired. In the Bayesian approach to inverse problems, edge-preservation is often achieved using Markov random field priors based on heavy-tailed distributions. Another strategy, popular in sparse statistics, is the application of hierarchical shrinkage priors. An advantage of this formulation lies in expressing the prior as a conditionally Gaussian distribution depending on heavy-tailed distributed hyper parameters. In this presentation, we revisit the shrinkage horseshoe prior and introduce its formulation for edge preserving settings. We discuss a Gibbs sampling framework to solve the Bayesian inverse problem. Applications from imaging science show that our computational procedure is able to compute sharp edge-preserving posterior point estimates with reduced uncertainty.
主讲人简介
董艺秋,丹麦科技大学副教授。 2007年于北京大学数学科学学院获得理学博士学位,师从徐树方教授和陈汉夫教授(岭南大学)。 2007年至2010年于奥地利格拉茨大学从事博士后研究。2010年至2013年于德国国家健康环境研究所做研究员。2013年加入丹麦科技大学为助理教授,于2015年转副教授。其研究方向主要集中于反问题数值计算、数值代数和数值最优化等数学领域,提出了若干解决这些应用问题的新的数学模型和数值算法。