Student Seminars

Dao Nguyen
PhD student, Statistics, University of Michigan

A second order iterated smoothing

“Plug and play” inferences, also known as “derivative free” or “likelihood free” inferences are receiving great attention recently due to the fact that in many practical problems, the likelihood is intractable to compute directly. This paper introduce a new plug and play algorithms, namely, second-order iterated smoothing. We use fixed lag smoothing to reduce the high variance of the estimator. To improve the speed of convergence, an approximation of the observed information matrix is also proposed. While enjoying greater convergence rate, most observed information matrix approximation methods are computational expensive, especially in plug and play approaches. Therefore, to relax the intensive computation, we propose a light-weight computational estimation of the Hessian. In addition, we propose to use random walk noise instead of independent white noise to efficiently explore the likelihood surface, increasing the convergence rate. Due to the special structure of iterated smoothing, the variances of the estimators are systematically controlled by kernel shrinkage. In a toy example and in a real data set, we show our proposed approaches with the same computational resources outweigh the standard approaches in term of convergence rate.

Student Seminar Archive

For questions regarding the Statistics Student Seminar or if you are interested in presenting, please contact Joonha Park(joonhap@umich.edu) or Jingshen Wang(jshwang@umich.edu).