I am working with Prof. Feng Liang and Prof. Jingbo Liu on the risk characterization in high-dimensional models. Inspired by the double descent phenomenon, we are interested in deriving the exact risk function for the estimator based on Zellner’s g-prior in high-dimensional models. We develop a methodology to select the optimal tuning parameter. We are also interested in accelerated sampling algorithms. Mainly using the higher-order Langevin dynamics, we want to achieve better convergence rates for corresponding MCMC algorithms.