Time: 4:30-5:30 p.m.
Date: Wednesday, October 29, 2025
Speaker: Xiao Wang, Head and J.O. Berger and M.E. Bock Professor of Statistics at Purdue University
Title: Neural Amortized Bayesian Computation
Abstract: Given an observation, how can we perform Bayesian inference when the likelihood is intractable? Approximate Bayesian Computation (ABC) provides a simple and intuitive solution with exactness guarantees. However, its performance deteriorates rapidly as the parameter dimension increases, a limitation known as the curse of dimensionality. We propose Neural Amortized Bayesian Computation (NABC), an extension of the ABC framework that alleviates this issue significantly while preserving its structure and exactness. NABC employs deep neural networks to capture and remove the dependence between data and parameters for efficient inference. With such exactness, this approach not only outperforms state-of-the-art simulation-based inference methods, such as Neural Posterior Estimation and Neural Likelihood Estimation, in amortized settings, but also achieves these improvements at substantially lower computational cost. We demonstrate its effectiveness on a range of benchmark studies reflecting practical scenarios and further extend its use to exact Bayesian inference for differentially privatized data.
Bio: Dr. Xiao Wang is Head and J.O. Berger and M.E. Bock Professor of Statistics at Purdue University. He earned his Ph.D. from the University of Michigan, and his research centers on machine learning, nonparametric statistics, and functional data analysis with particular emphasis on developing methods for high-dimensional and complex data. His work has been featured in leading statistical journals and machine learning conferences, and he is a fellow of the Institute of Mathematical Statistics (IMS) and the American Statistical Association (ASA). He currently serves as an associate editor for JASA, Technometrics, and Lifetime Data Analysis.