3月28日贺诗源(中国人民大学)报告

发布者:周晓英发布时间:2019-03-19浏览次数:572

统计与数据科学学院

---------------------------------------

学术讲座

讲座题目:Simultaneous estimation of periods and period-luminosity relations for multi-band Miras via stochastic variational inference

人:贺诗源助理教授

        中国人民大学统计与大数据研究院

时间:2019328(周四)上午10:30

地点:统计与数据科学学院126教室

摘要:

Miras are periodical stars whose one cycle of brightness variation lasts more than a hundred days. Their period-luminosity relation has become a promising tool for distance measurement in astronomy. For sparsely sampled multi-band light curves of a Mira, the task of identifying the correct period using all available information is challenging, as the posterior distribution of its period is usually highly multi-modal. We develop a hierarchical Bayesian model to simultaneously estimate the periods for a collection of Miras, as well as their period-luminosity relation. The hierarchical model exploits the power of joint modeling and use the information from all samples to improve estimation accuracy. With stochastic variational inference, the model inference is cast into an optimization problem, which is scalable to large scale modern astronomical surveys. An efficient algorithm is developed to update all components of the variational distributions. Through simulation studies, we demonstrate the proposed method effectively improves period estimation accuracy. Finally, the proposed method is applied to Miras in M33 and an estimation of distance modulus is obtained.

邀请人:张博海博士

欢迎广大师生参加!

统计与数据科学学院