Multidimensional normal approximation by Stein's method and Bismut formula
Speaker :
Time :
Location :
Title: Multidimensionalnormal approximation by Stein's method and Bismut formula
Speaker: Dr. Lihu Xu (徐礼虎), 澳门大学
Time: 3:00pm, 2018-12-21
Location: 200-9 The2th Floor, Sir R. R. Shaw Building,
Yuquan Campus, Zhejiang University
Abstract: Stein's method has been widely used for probabilityapproximations. However, in the multi-dimensional setting, most of the resultsare for multivariate normal approximation or for test functions with boundedsecond- or higher-order derivatives. For a class of multivariate limitingdistributions, we use Bismut's formula in Malliavin calculus to control thederivatives of the Stein equation solutions by the first derivative of the testfunction.Combined with Stein's exchangeable pair approach, we obtain a generaltheorem for multivariate approximations with near optimal error bounds on theWasserstein distance. We apply thetheorem to the unadjusted Langevin algorithm.
All are welcome!
Contact Person:Zhonggen Su, suzhonggen@zju.edu.cn
浙江大学统计研究所
2018-12-10