Mean-Field Approximations for High-Dimensional Bayesian Regression

Watch a research lecture from Harvard CMSA where Subhabrata Sen explores mean-field approximations in high-dimensional Bayesian regression. Dive into the mathematical analysis of Naive Mean-Field (NMF) approximations as alternatives to MCMC-based methods for posterior distribution estimation in Bayesian inference. Learn about recent developments in understanding NMF accuracy under structural constraints and examine scenarios where advanced mean-field techniques like Bethe approximation prove more effective. The presentation covers collaborative research findings with Sumit Mukherjee from Columbia University and Jiaze Qiu from Harvard University, offering insights into the duality of approximation methods in high-dimensional settings.